PRE2016 4 Groep1: Difference between revisions

From Control Systems Technology Group
Jump to navigation Jump to search
 
(16 intermediate revisions by 2 users not shown)
Line 6: Line 6:
* Dennis Struver 0955477
* Dennis Struver 0955477


= Project definition =
= Project Definition =
In this chapter the project is defined by elaborating the subject, goal, approach, objectives, deliverables and the project planning.
In this chapter the project is defined by elaborating the subject, goal, approach, objectives, deliverables and the project planning.
   
   
Line 290: Line 290:
{|style="margin: none;"  
{|style="margin: none;"  
| [[File:cause+agegroup.PNG|thumb|none|200 px|Figure 2: Number of fatal accidents for each age group]]
| [[File:cause+agegroup.PNG|thumb|none|200 px|Figure 2: Number of fatal accidents for each age group]]
| [[File:2008accidents.PNG|thumb|300 px|Table 3: Number of home-related accidents for 2000-2008]]
| [[File:2008accidents.PNG|thumb|300 px|Table 4: Number of home-related accidents for 2000-2008]]
|}
|}
   
   
From 2000 to 2008, there was an annual average of 30.569 unintentional injury deaths occurring in the home environment in the USA. This number is almost twice as high as in 1999 where the annual unintentional home injury deaths were just more than 18.000.
In Table 4 it is shown that from 2000 to 2008, there was an annual average of 30.569 unintentional injury deaths occurring in the home environment in the USA. This number is almost twice as high as in 1999 where the annual unintentional home injury deaths were just more than 18.000.
In Figure 2 it becomes clear that the most unintentional injury deaths are caused by poisoning, fall, fire/burn and choking/suffocation. This does show that these causes are also the biggest issues from the year 2000 to 2008. In Table 4 the age-adjusted rate of unintentional home injury deaths for 2000-2008 is shown.
In Figure 2 it becomes clear that the most unintentional injury deaths are caused by poisoning, fall, fire/burn and choking/suffocation. This does show that these causes are also the biggest issues from the year 2000 to 2008. In Table 5 the age-adjusted rate of unintentional home injury deaths for 2000-2008 is shown.


{|style="margin: none;"
{|style="margin: none;"
| [[File:2008age.PNG|thumb|600 px|Table 4: Age-adjusted rate of unintentional home injury death’s]]
| [[File:2008age.PNG|thumb|600 px|Table 5: Age-adjusted rate of unintentional home injury death’s]]
| [[File:Table2EUR.PNG|thumb|none|600x271px|Table 5: Home injury deaths in 16 European countries, annual number, percentage, age 0-14, all causes, 2002-2004]]
| [[File:Table2EUR.PNG|thumb|none|600x271px|Table 6: Home injury deaths in 16 European countries, annual number, percentage, age 0-14, all causes, 2002-2004]]
|}  
|}  


The 3 major causes of unintentional home injury death of the age group 0 to 9 are still suffocation, drowning and fire/burn, which is the same result as the research from 1992 to 1999. It is clear that the fatalities inside a house for children stay about the same looking at the results from 1992 to 2008. Therefore it is assumed that these causes are nowadays still present and that the most fatal injuries are caused by suffocation, drowning and fire or burns.
The 3 major causes of unintentional home injury death of the age group 0 to 9 are still suffocation, drowning and fire/burn, which is the same result as the research from 1992 to 1999. It is clear that the fatalities inside a house for children stay about the same looking at the results from 1992 to 2008. Therefore it is assumed that these causes are nowadays still present and that the most fatal injuries are caused by suffocation, drowning and fire or burns.
   
   
The leading mechanisms of death from unintentional injury in children are road traffic crashes, drowning, poisoning, thermal injuries and falls. The top four causes of unintentional fatal home injuries in children 0–14 years in all countries aggregated were drowning/submersion, fire/flames, poisoning and falls, see Table 5. These causes accounted for almost 75% of all home injury deaths<ref>http://www.euro.who.int/__data/assets/pdf_file/0003/83757/E92049.pdf</ref>
The leading mechanisms of death from unintentional injury in children are road traffic crashes, drowning, poisoning, thermal injuries and falls. The top four causes of unintentional fatal home injuries in children 0–14 years in all countries aggregated were drowning/submersion, fire/flames, poisoning and falls, see Table 6. These causes accounted for almost 75% of all home injury deaths<ref>http://www.euro.who.int/__data/assets/pdf_file/0003/83757/E92049.pdf</ref>


===Unintentional non-fatal home injuries===
===Unintentional non-fatal home injuries===
   
   
The non-fatal injury rate for children younger than 9 years old In the USA was 28.054 per 100.000 in 2000-2006. For this age group falls accounted for the largest amount of injuries. In Table 6 the top five leading causes of unintentional injuries in the USA are displayed. Getting struck by or against an object was the second largest cause of injuries. For children aged 1-4 cuts or pierce injuries account for 4% of the total amount of injuries (compared to 43% for falls) and for children aged 5-9 cuts or pierce injuries account for 7% of the total amount (compared to 37% for falls). When children grow older outside related accidents (transportation) become more important.
The non-fatal injury rate for children younger than 9 years old In the USA was 28.054 per 100.000 in 2000-2006. For this age group falls accounted for the largest amount of injuries. In Table 7 the top five leading causes of unintentional injuries in the USA are displayed. Getting struck by or against an object was the second largest cause of injuries. For children aged 1-4 cuts or pierce injuries account for 4% of the total amount of injuries (compared to 43% for falls) and for children aged 5-9 cuts or pierce injuries account for 7% of the total amount (compared to 37% for falls). When children grow older outside related accidents (transportation) become more important.
   
   
{|style="margin: none;"
{|style="margin: none;"
| [[File:Tabel1 CDC.JPG|thumb|none|600x246px|Table 6: Top Five Leading Causes of Unintentional Injury Deaths and Nonfatal Injuries among Children 0 to 19 Years, by Age Group, United States, 2000–2006 ]]
| [[File:Tabel1 CDC.JPG|thumb|none|600x246px|Table 7: Top Five Leading Causes of Unintentional Injury Deaths and Nonfatal Injuries among Children 0 to 19 Years, by Age Group, United States, 2000–2006 ]]
| [[File:Tabel2 unintentional injury.JPG|thumb|none|600x271px|Table 7: National estimates of number of nonfatal, unintentional injuries, by location, United States, 1998–1999]]
| [[File:Tabel2 unintentional injury.JPG|thumb|none|600x271px|Table 8: National estimates of number of nonfatal, unintentional injuries, by location, United States, 1998–1999]]
|}
|}
   
   
Line 317: Line 317:
   
   
A study conducted in the late 1990’s also showed an alarming number of accidents that happen in the home environment.
A study conducted in the late 1990’s also showed an alarming number of accidents that happen in the home environment.
The NHAMCS data indicate that nearly 9.8 million emergency department and 1.4 million outpatient hospital visits were made in 1999 for nonfatal, unintentional injuries that took place in a home environment as shown in Table 7. Likewise, data obtained in the NHIS include 12.922.220 nonfatal unintentional home injuries, excluding poisonings, requiring some form of medical advice. NHIS data also show that 750.000 persons aged 5 years were reported as missing at least 1 day of school, as a result of an unintentional home injury.
The NHAMCS data indicate that nearly 9.8 million emergency department and 1.4 million outpatient hospital visits were made in 1999 for nonfatal, unintentional injuries that took place in a home environment as shown in Table 8. Likewise, data obtained in the NHIS include 12.922.220 nonfatal unintentional home injuries, excluding poisonings, requiring some form of medical advice. NHIS data also show that 750.000 persons aged 5 years were reported as missing at least 1 day of school, as a result of an unintentional home injury.
   
   
Each data set identified falls as the most common mechanism of injury by far, accounting for 36.2% to 45.7% of the injuries or visits to healthcare providers
Each data set identified falls as the most common mechanism of injury by far, accounting for 36.2% to 45.7% of the injuries or visits to healthcare providers
Line 365: Line 365:
'''Pressure Fit Baby Gates '''  
'''Pressure Fit Baby Gates '''  


In the past they were simply barriers than were wedged between two walls or in a doorway and you had to step over them in order to go through or remove them completely. Much has changed. Nowadays the gate itself stays across the opening and is held in place by the pressure that is usually created by extending threaded pressure pads to the wall or door jam, these gates have a door that can be opened and closed for convenience. Because this type has a gate within a gate it also has a threshold across the bottom when the gate is opened. Two things all pressure gates have in common is that they require two flat surfaces across from each other to be mounted against and they cannot be mounted on an angle. These gates generally are not designed to be mounted on staircases.
In the past they were simply barriers that were wedged between two walls or in a doorway and you had to step over them in order to go through or remove them completely. Much has changed. Nowadays the gate itself stays across the opening and is held in place by the pressure that is usually created by extending threaded pressure pads to the wall or door jam, these gates have a door that can be opened and closed for convenience. Because this type has a gate within a gate it also has a threshold across the bottom when the gate is opened. Two things all pressure gates have in common is that they require two flat surfaces across from each other to be mounted against and they cannot be mounted on an angle. These gates generally are not designed to be mounted on staircases.
   
   
'''Hardware Mounted Baby Gates '''  
'''Hardware Mounted Baby Gates '''  
Line 388: Line 388:
[[File:Hallways.PNG|thumb|600x264px|Figure 5: Possible hallways.]]
[[File:Hallways.PNG|thumb|600x264px|Figure 5: Possible hallways.]]


To target the largest group of potential customers, the system should be able to operate at as much different house settings as possible. The staircase itself can deffer per house and the system should be able to be implemented in every house. The design should be applicable for every type of stairs. For example U-shaped stairs, normal straight stairs or curved stairs.
To target the largest group of potential customers, the system should be able to operate at as much different house settings as possible. The staircase itself can differ per house and the system should be able to be implemented in every house. The design should be applicable for every type of stairs. For example U-shaped stairs, normal straight stairs or curved stairs.
   
   
''' Hallways '''
''' Hallways '''
Line 411: Line 411:
The system must be able to work with the information it gets from the detection and recognition software. The information that is acquired must be processed and converted into actions. This processing and converting must be done almost immediately. Therefore the system must be able to work together with multiple components at the same time.
The system must be able to work with the information it gets from the detection and recognition software. The information that is acquired must be processed and converted into actions. This processing and converting must be done almost immediately. Therefore the system must be able to work together with multiple components at the same time.


= Recognition and detection systems =
= Recognition and Detection Systems =
For the system to work properly, a proper detection and localization system has to be chosen. The system has to know if there is a person in the defined space and to know its location to at when necessary. It is also necessary for the detection system to be able to differentiate between children and adults because the system has to act when a parent is near the gate. A few different detection systems have been investigated and compared to choose the best one. There is also looking to the set requirements to make sure the system satisfy these requirements.  
For the system to work properly, a proper detection and localization system has to be chosen. The system has to know if there is a person in the defined space and to know its location to at when necessary. It is also necessary for the detection system to be able to differentiate between children and adults because the system has to act when a parent is near the gate. A few different detection systems have been investigated and compared to choose the best one. There is also looking to the set requirements to make sure the system satisfy these requirements.  


Line 420: Line 420:
One way to locate an object is by making use of infrared (IR) sensors. Simple IR sensors are widely used nowadays in robotics and automation, process control, remote sensing, and safety and security systems. More specifically, they have been used in object and proximity detection, counting, distance and depth monitoring, floor sensing, position measurement and control, obstacle/collision avoidance, and map building. Since the great variety of applications of IR, IR sounds promising for the detection of a person and the differentiating between adults and children.  
One way to locate an object is by making use of infrared (IR) sensors. Simple IR sensors are widely used nowadays in robotics and automation, process control, remote sensing, and safety and security systems. More specifically, they have been used in object and proximity detection, counting, distance and depth monitoring, floor sensing, position measurement and control, obstacle/collision avoidance, and map building. Since the great variety of applications of IR, IR sounds promising for the detection of a person and the differentiating between adults and children.  


IR is also commonly used for face recognition systems, most of the time for safety reasons. Thermal face recognition deals with the face recognition system that takes the thermal heat of the face as an input. Thermal human face images are generated due to the thermal human body heat. Such a face recognition system would be also applicable for the differentiating a child from an adult. The generated thermal human face image of the observed person could be compared with an image of the child out of a database and in this way the system is able to differentiate the two. A downside of such a system is that the resolution of the thermal images is not very high. Also is it very plausible that the child will not always look straight into the camera, so different face position and also facial expressions should be covered by the system These downsides together with the high cost of such high-end IR cameras make this application of IR not interesting for this application. <ref>Mrinal Kanti B, Thermal infrared face recognition - a biometric identification technique for robust security system, Intech, 2011 </ref>
IR is also commonly used for face recognition systems, most of the time for safety reasons. Thermal face recognition deals with the face recognition system that takes the thermal heat of the face as an input. Thermal human face images are generated due to the thermal human body heat. Such a face recognition system would be also applicable for the differentiating a child from an adult. The generated thermal human face image of the observed person could be compared with an image of the child out of a database and in this way the system is able to differentiate the two. A downside of such a system is that the resolution of the thermal images is not very high. Also is it very plausible that the child will not always look straight into the camera, so different face position and also facial expressions should be covered by the system. These downsides together with the high cost of such high-end IR cameras make this application of IR not interesting for this application. <ref>Mrinal Kanti B, Thermal infrared face recognition - a biometric identification technique for robust security system, Intech, 2011 </ref>


Since the costs of such a system, to be able to cope with the state-of-the-art, have to be as low as possible, the localization system should not be too expensive. Therefore the applications of low-cost IR sensors is investigated. To still be able to differentiate objects, the differentiating techniques employed should be different from such operations performed on conventional images. The targets which have to be differentiated are not patterns in a 2D imaged, but rather objects in space, exhibiting depth and at different positions with respect to the sensing system. This would be a common situation which often will occur when the system is in use in a SMART home.
Since the costs of such a system, to be able to cope with the state-of-the-art, have to be as low as possible, the localization system should not be too expensive. Therefore the applications of low-cost IR sensors is investigated. To still be able to differentiate objects, the differentiating techniques employed should be different from such operations performed on conventional images. The targets which have to be differentiated are not patterns in a 2D imaged, but rather objects in space, exhibiting depth and at different positions with respect to the sensing system. This would be a common situation which often will occur when the system is in use in a SMART home.


There are different techniques one can use to differentiate the geometry of an object with such low-cost IR sensors. Here only the template based approach to differentiate only the geometry of the target object is discussed, since only the geometry is important for this application. The template-based approach is based on comparing the acquired IR intensity scans with previously stored templates acquired from targets with different properties. Hence, this approach relies on the distinctive natures of the IR intensity scans and requires the storage of a complete set of reference scans of interest. For targets made of the same material, but with different geometrical properties, the correct differentiation rate is 97 percent.
There are different techniques one can use to differentiate the geometry of an object with such low-cost IR sensors. Here, only the template based approach to differentiate only the geometry of the target object is discussed, since only the geometry is important for this application. The template-based approach is based on comparing the acquired IR intensity scans with previously stored templates acquired from targets with different properties. Hence, this approach relies on the distinctive natures of the IR intensity scans and requires the storage of a complete set of reference scans of interest. For targets made of the same material, but with different geometrical properties, the correct differentiation rate is 97 percent.


The geometry of children and adults is a major difference since a child is much smaller than an average adult. Therefore only a rough estimation of the size of the object is enough to differentiate a child from an adult. The geometry itself is therefore not important, one could approximate the person as a solid block or cylinder with specific dimensions of the length, width, and depth. To determine the full geometry of a person is also way too complex for such a low-cost sensor. Since it is possible to determine the size of an object, this technique would be suitable for this application. <ref>Barshan B. Taget differentiation with simple infrared sensors using statistical pattern recognition techniques, ScienceDirect, 22 March 2006 </ref>
The geometry of children and adults is a major difference since a child is much smaller than an average adult. Therefore only a rough estimation of the size of the object is enough to differentiate a child from an adult. The geometry itself is therefore not important, one could for example approximate the persons geometry as a solid block or cylinder with specific dimensions of the length, width, and depth. To determine the full geometry of a person is also way too complex for such a low-cost sensor. Since it is possible to determine the size of an object, this technique would be suitable for this application. <ref>Barshan B. Taget differentiation with simple infrared sensors using statistical pattern recognition techniques, ScienceDirect, 22 March 2006 </ref>


So far this system sounds very promising. The sensors used are of low cost, it is able to detect whether there is a person in the room and the differentiation of the geometry is of great accuracy. There are only a couple remarks to make.  
So far this system sounds very promising. The sensors used are of low cost, it is able to detect whether there is a person in the room and the differentiation of the geometry is of great accuracy. There are only a couple remarks to make.  


The system determines the geometry by sensing the object, therefore the sensor should be mounted on a gimbal which can rotate the sensor in all directions. IF the sensor is not able to rotate, it will not be able to fully sense the target and detection and differentiation becomes impossible.
The system determines the geometry by sensing the object, therefore the sensor should be mounted on a gimbal which can rotate the sensor in all directions. If the sensor is not able to rotate, it will not be able to fully sense the target and detection and differentiation becomes impossible.
Since the system can only determine simple geometries, objects as for example furniture can cause false positives and negative for this system. The system can track these objects and because the geometry of a cabinet can be the same as the modeled geometry of a child, the simplicity of this system can become a problem.  
Since the system can only determine simple geometries, objects as for example furniture can cause false positives and negatives for this system. The system can track these objects and because the geometry of a cabinet can be the same as the simple modeled geometry of a child, the simplicity of this system can become a problem.  
The last problem is that it cost time for the system to fully observe the target since the full geometry has to be sensed. Moving targets, therefore, become a problem, because a child that moves will not be fully scanned by the system and therefore detection and differentiation become impossible.  
The last problem is that it cost time for the system to fully observe the target since the full geometry has to be sensed. Moving targets, therefore, become a problem, because a child that moves will not be fully scanned by the system and therefore detection and differentiation become impossible.  
The simplicity of such low-cost IR sensing systems is preferable and promising, but due to the downsides of it as listed above, this way of using IR will not be suitable for this application.
The simplicity of such low-cost IR sensing systems is preferable and promising, but due to the downsides of it as listed above, this way of using IR will not be suitable for this application.
Line 441: Line 441:
Infrared can also be used in a different way. In the section above is discussed how IR sensors can be used to locate objects in an environment using the intensity of the reflected light. However, IR can also be used to locate objects by thermal mapping. There exist an indoor localization and monitoring system, which is based on a wireless and PIR (WPIR) sensor fusion system. The PIR sensor transforms incident IR radiation into an electrical signal. PIR detects changes in temperature coinciding with movement of a person or object in the environment. The output of the PIR sensor is in disorder for human movement detection. A human walking through a PIR sensor detecting region and the corresponding output signal is shown in Figure 6.  
Infrared can also be used in a different way. In the section above is discussed how IR sensors can be used to locate objects in an environment using the intensity of the reflected light. However, IR can also be used to locate objects by thermal mapping. There exist an indoor localization and monitoring system, which is based on a wireless and PIR (WPIR) sensor fusion system. The PIR sensor transforms incident IR radiation into an electrical signal. PIR detects changes in temperature coinciding with movement of a person or object in the environment. The output of the PIR sensor is in disorder for human movement detection. A human walking through a PIR sensor detecting region and the corresponding output signal is shown in Figure 6.  


WPIR has proven it can monitor multiple targets with relative good resolution. This is promising for the SMART house application, but it needs to be able to differentiate between adults and children. It is possible for WPIR to differentiate humans and robots. This because the signals the sensor receives are different when a robot passes the sensor and when a human does. Based on this, WPIR is able to differentiate. But since the signal of adults and parents probably will not differ much because the thermal properties are the same, WPIR needs some adjustments to be able to differentiate children from parents.
WPIR has proven it can monitor multiple targets with relative good resolution. This is promising for the SMART house application, but it needs to be able to differentiate between adults and children. It is possible for WPIR to differentiate humans and robots. This is because the signals the sensor receives are different when a robot passes the sensor and when a human does. Based on this, WPIR is able to differentiate. But since the signal of adults and parents probably will not differ much because the thermal properties are the same, WPIR needs some adjustments to be able to differentiate children from parents.


What is very promising of WPIR is that the implementation is very easy and the costs are also low. The only hardware needed is a ceiling camera which can observe the environment. Since it is also possible to monitor multiple targets WPIR can still be suitable for this application, but then the problem of differentiation should be solved. <ref>R. C. Luo and O. Chen, "Wireless and Pyroelectric Sensory Fusion System for Indoor Human/Robot Localization and Monitoring," in IEEE/ASME Transactions on Mechatronics, vol. 18, no. 3, pp. 845-853, June 2013. doi: 10.1109/TMECH.2012.2188300</ref>
What is very promising of WPIR is that the implementation is very easy and the costs are also low. The only hardware needed is a ceiling camera which can observe the environment. Since it is also possible to monitor multiple targets WPIR can still be suitable for this application, but then the problem of differentiation should be solved. <ref>R. C. Luo and O. Chen, "Wireless and Pyroelectric Sensory Fusion System for Indoor Human/Robot Localization and Monitoring," in IEEE/ASME Transactions on Mechatronics, vol. 18, no. 3, pp. 845-853, June 2013. doi: 10.1109/TMECH.2012.2188300</ref>
Line 449: Line 449:
The ultrasound system is not always on the point of view of the today technologies, but there is no reason for this. Even nature shows that the ultrasound detection system works fine in any types of conditions. How exactly does it work? Bats have this power to orientate themselves only by using ultrasound waves that at the contact with objects return to the source. Based on the time of such a routine, the bat can detect exactly how far the target is.
The ultrasound system is not always on the point of view of the today technologies, but there is no reason for this. Even nature shows that the ultrasound detection system works fine in any types of conditions. How exactly does it work? Bats have this power to orientate themselves only by using ultrasound waves that at the contact with objects return to the source. Based on the time of such a routine, the bat can detect exactly how far the target is.


This feature sounds good and it seems to work, but there is a big counter argument for using it. Especially in the case that is being addressed in this project, where the subjects are young children. This method is an invasive one, which in the long run can cause a lot of problems, especially because of its chemical effects. As stated in “The chemical effects of ultrasound” <ref>http://www.scs.illinois.edu/suslick/documents/sciamer8980.pdf</ref>, it can drive metal particles together at such high speeds that they melt at the point of collision, and ultrasound can generate microscopic flames in cold liquids. As it is already known, in blood there are different metals like Potassium, Iron, Calcium, Mercury, Sodium and many others which can react to ultrasound waves. Effects that can occur are related to headache, dizziness, and nausea, but most important one is the hearing damage. So far, the most difficult part of using such a technology is the long time exposure of subjects to the system and the age range of the subject is below 9 years. Young people are more sensitive to any types of factors compared to adults, hence it is difficult to integrate this technology into the current project analysis. The same document stated that the technology is used for industrial applications because it can radiate through large volumes of liquid, it will represent a barrier instead of a helping tool for the SMART home system. Since due to this downside ultrasound is not an option, this detection method is not further analyzed.  
This feature sounds good and it seems to work, but there is a big counter argument for using it. Especially in the case that is being addressed in this project, where the subjects are young children. This method is an invasive one, which in the long run can cause a lot of problems, especially because of its chemical effects. As stated in “The chemical effects of ultrasound”, it can drive metal particles together at such high speeds that they melt at the point of collision, and ultrasound can generate microscopic flames in cold liquids. As it is already known, in blood there are different metals like potassium, iron, calcium, mercury, sodium and many others which can react to ultrasound waves. Effects that can occur are related to headache, dizziness, and nausea, but most important one is the hearing damage. So far, the most difficult part of using such a technology is the long time exposure of subjects to the system and the age range of the subject is below 9 years. Young people are more sensitive to any types of factors compared to adults, hence it is difficult to integrate this technology into the current project analysis. The same document stated that the technology is used for industrial applications because it can radiate through large volumes of liquid, it will represent a barrier instead of a helping tool for the SMART home system. Since due to this downside ultrasound is not an option, this detection method is not further analyzed. <ref>http://www.scs.illinois.edu/suslick/documents/sciamer8980.pdf</ref>


''' Pressure Sensor '''
''' Pressure Sensor '''
Line 460: Line 460:


The active badge is a badge that a human wears so that he/she can be tracked by a sensing system. The Active Badge operates as a beacon, regularly signaling a unique code to a network of sensors distributed around the area to be monitored. Sightings are gathered by using a master processor which polls the sensors through a network provided for the purpose. The name and location of a badge wearer can be ascertained by looking up the badge ID in a table and looking up the location where the sighting was made. It has been continually improved on to make it more accurate. Because of these improvements, this technology is applicable for the SMART home, because it is able to track multiple people, even through walls and it is not expensive to implement. Since every user has its own ID, differentiating between adults and children should not be a problem.  
The active badge is a badge that a human wears so that he/she can be tracked by a sensing system. The Active Badge operates as a beacon, regularly signaling a unique code to a network of sensors distributed around the area to be monitored. Sightings are gathered by using a master processor which polls the sensors through a network provided for the purpose. The name and location of a badge wearer can be ascertained by looking up the badge ID in a table and looking up the location where the sighting was made. It has been continually improved on to make it more accurate. Because of these improvements, this technology is applicable for the SMART home, because it is able to track multiple people, even through walls and it is not expensive to implement. Since every user has its own ID, differentiating between adults and children should not be a problem.  
A disadvantage is that the users will have to wear these badges which might be considered inconvenient by adults. On the other hand, a child might not wear it, for example when the parent forgets to put it on, or even destroy it while wearing it. Also if the badge is attached to a piece of clothing is the child able to remove that clothing. If a child is not wearing the badge, the system will not be able to track the child and the same as with the pressure sensor, the system does not oversee the full environment. These downsides bring the accuracy down to this detection system since there are a lot of bugs which can let the detection fail. <ref>R. Want and A. Hopper, "Active badges and personal interactive computing objects," in IEEE Transactions on Consumer Electronics, vol. 38, no. 1, pp. 10-20, Feb 1992. doi: 10.1109/30.125076 </ref><ref>Y. Zhao, N. Patwari, P. Agrawal and M. Rabbat, "Directed by Directionality: Benefiting from the Gain Pattern of Active RFID Badges," in IEEE Transactions on Mobile Computing, vol. 11, no. 5, pp. 865-877, May 2012. doi: 10.1109/TMC.2011.89</ref>
A disadvantage is that the users will have to wear these badges which might be considered inconvenient by adults. On the other hand, a child might not wear it, for example when the parent forgets to put it on, or even destroy it while wearing it. Also if the badge is attached to a piece of clothing is the child able to remove that clothing. If a child is not wearing the badge, the system will not be able to track the child and the same as with the pressure sensor, the system does not oversee the full environment. These downsides bring the accuracy down to this detection system since there are a lot of bugs which can let the detection fail. <ref>R. Want and A. Hopper, "Active badges and personal interactive computing objects," in IEEE Transactions on Consumer Electronics, vol. 38, no. 1, pp. 10-20, Feb 1992. doi: 10.1109/30.125076 </ref><ref>Y. Zhao, N. Patwari, P. Agrawal and M. Rabbat, "Directed by Directionality: Benefiting from the Gain Pattern of Active RFID Badges," in IEEE Transactions on Mobile Computing, vol. 11, no. 5, pp. 865-877, May 2012. doi: 10.1109/TMC.2011.89</ref>


''' Camera Detection '''
''' Camera Detection '''


Another detecting system is camera vision. This alternative represents the best one for indoor use for detecting people. It is not invasive, which means it can also be used to detect children, it is rather cheap and efficient. But what is the state-of-the-art of such a technology? The latest cameras are connected to the internet of things which make them SMART. They are able to process via an AI all the given images and provide helpful data. The bad side of it is represented by the privacy issues, but looking at the SMART house application which is the underlying thought of this project, this would not be a problem since such a house consists of multiple sensors.  
Another detecting system is camera vision. This alternative represents the best one for indoor use for detecting people. It is not invasive, which means it can also be used to detect children, it is rather cheap and efficient. But what is the state-of-the-art of such a technology? The latest cameras are connected to the internet of things which make them SMART. They are able to process via an AI all the given images and provide helpful data. The down side of it is represented by the privacy issues, but looking at the SMART house application which is the underlying thought of this project, this would not be a problem since such a house consists of multiple sensors.  


Camera systems like ‘Kinect’ are able to track and observe people. This is done by software which analyzes the camera image of the Kinect camera system. This is done as showed in Figure 7.
Camera systems like ‘Kinect’ are able to track and observe people. This is done by software which analyzes the camera image of the Kinect camera system. This is done as showed in Figure 7.
Line 470: Line 471:
[[File:DetectionCamera.jpg|300px|thumb|Figure 7: Detection and tracking with camera vision]]
[[File:DetectionCamera.jpg|300px|thumb|Figure 7: Detection and tracking with camera vision]]


A person detector which makes use of such Kinect camera data or RGB-D is HOG (Histograms of oriented gradients). HOG is nowadays one of the most performant and widely used methods for visual people detection. The method considers a fixed-size detection window which is densely subdivided into a uniform grid of cells. For each cell, the gradient orientations over the pixels are computed and collected in a 1D histogram. The intuition is that local appearance and shape can be characterized by a distribution of local gradients without the precise knowledge of their position in the cell. <ref>L. Spinello, "People Detection in RGB_D Data", in  International Conference on Intelligent Robots and Systems, September 2011</ref>
A person detector which makes use of such Kinect camera data or RGB-D is HOG (Histograms of Oriented Gradients). HOG is nowadays one of the most performant and widely used methods for visual people detection. The method considers a fixed-size detection window which is densely subdivided into a uniform grid of cells. For each cell, the gradient orientations over the pixels are computed and collected in a 1D histogram. The intuition is that local appearance and shape can be characterized by a distribution of local gradients without the precise knowledge of their position in the cell. <ref>L. Spinello, "People Detection in RGB_D Data", in  International Conference on Intelligent Robots and Systems, September 2011</ref>


With the HOG and the RGB-D, an image can be created where a person is detected and marked with a colored box. The height of this box can be determined and in this way, adults and children can be differentiated from each other. Because the height difference between an adult and a child is big enough, the accuracy of the height of the box is not that important since the box of the adult will be way bigger than the one of the child. Therefore, the accuracy whether a person is detected or not is more important than the accuracy of the box itself. The performance of the people detection system is evaluated in terms of detection rates (accuracy) and false positives/negatives. True positives are the people images detected from the ground truth. False negatives are the people images not detected from the ground truth and false positives are images detected as people that do not appear in the ground truth. It is logical that the system will fail when the adult is not detected by the system but wants to go down the stairs, a false negative. Also, dangerous situations can appear when a false positive is detected and the child has a free entrance to the stairs. The accuracy of the detection system is then defined as follows:
With the HOG and the RGB-D, an image can be created where a person is detected and marked with a colored box. The height of this box can be determined and in this way adults and children can be differentiated from each other. Because the height difference between an adult and a child is big enough, the accuracy of the height of the box is not that important since the box of the adult will be way bigger than the one of the child. Therefore, the accuracy whether a person is detected or not is more important than the accuracy of the box itself. The performance of the people detection system is evaluated in terms of detection rates (accuracy) and false positives/negatives. True positives are the people images detected from the ground truth. False negatives are the people images not detected from the ground truth and false positives are images detected as people that do not appear in the ground truth. It is logical that the system will fail when the adult is not detected by the system but wants to go down the stairs, a false negative. Also, dangerous situations can appear when a false positive is detected and the child has a free entrance to the stairs. The accuracy of the detection system is then defined as follows:
    
    
<math>Accuracy=\frac{TP+TN}{TP+FP+TN+FN}</math>
<math>Accuracy=\frac{TP+TN}{TP+FP+TN+FN}</math>
Line 482: Line 483:
Another advantage of using thermal images instead of normal camera view is that a normal camera fails when operating in the dark. Persons cannot be detected when the camera has lost all his vision. To cope with this problem again thermal images comes in handy, a thermal camera will not loose its functions when the environment is dark since heat is always generated by the targets.  
Another advantage of using thermal images instead of normal camera view is that a normal camera fails when operating in the dark. Persons cannot be detected when the camera has lost all his vision. To cope with this problem again thermal images comes in handy, a thermal camera will not loose its functions when the environment is dark since heat is always generated by the targets.  


The second problem of normal cameras is that the precise location of the target is not known. The detection system uses the 2D image to detect the targets, this can cause some problems. For example, imagine a child in front of the image and an adult in the back. When the detection system does his work, the two boxes may be of the same height because the parent is much smaller in a 2D image when it is positioned further back. The system, therefore, lost its property to differentiate properly. To still know the exact location and height of a target and prevent false positives from happening, a device which measures distances is necessary for this system. This makes the system more expensive, but the accuracy will go up together with the reliability. When the distance is known and the height of the box, the software of the system can do the rest by calculating the exact height and is able to differentiate. Also, the system can keep track of the target and see when it is near the stairs so it can act when necessary. Later more will be elaborated on this distance sensor.   
The second problem of normal cameras is that the precise location of the target is not known. The detection system uses the 2D image to detect the targets, this can cause some problems. For example, imagine a child in front of the image and an adult in the back. When the detection system does his work, the two boxes may be of the same height because the parent is much smaller in a 2D image when it is positioned further back. The system, therefore, lost its property to differentiate properly. To still know the exact location and height of a target and prevent false positives from happening, a device which measures distances is necessary for this system. This makes the system more expensive, but the accuracy will go up together with the reliability. When the distance is known and the height of the box, the software of the system can do the rest by calculating the exact height of the person and it is able to differentiate. Also, the system can keep track of the target and see when it is near the stairs so it can act when necessary. Later more will be elaborated on this distance sensor.   


''' Thermal Imaging '''
''' Thermal Imaging '''


Thermal cameras are a really good way to detect people. They are often used by firefighters to detect the source of the fire or see the heat signatures of the casualties. This detection system is already in use and it was proved to be efficient, which may make it worth implementing in a SMART house.  
Thermal cameras are a really good way to detect people. They are often used by firefighters to detect the source of the fire or see the heat signatures of the casualties. This detection system is already in use and it was proved to be efficient, which may make it worth implementing in a SMART house.  
The main disadvantage is constituted by the fact that the images appear like 2D plane pictures, where the actual distances to objects are not known. For a real case situation where people can use these images to spot other persons thermal cameras perform well, but how do they cope with human detection when heat objects are around? There might occur several problems with shape detection when applied to a SMART home system for children safety. The thermal imaging shows the details of a person usually in colors of red, yellow and orange. These colors are also shown for hot objects, for example, a cup of coffee.  
The main disadvantage is constituted by the fact that the images appear like 2D plane pictures, where the actual distances to objects are not known as stated above. For a real case situation where people can use these images to spot other persons thermal cameras perform well, but how do they cope with human detection when heat objects are around? There might occur several problems with shape detection when applied to a SMART home system for children safety. The thermal imaging shows the details of a person usually in colors of red, yellow and orange. These colors are also shown for hot objects, for example, a cup of coffee.  


Thus, the main problem for the AI which shall detect persons is how exactly to distinguish between a cup of hot liquid which is close to the camera and a person which is far behind? Since the shapes are not perfectly sharp, there might be used a human body template to look for persons in such images, but another question arises: how do you identify between two persons that are on the same line facing the camera? The image will show a bunch of heat and the AI may identify one person, but it is impossible to find the second person just by using that camera. So, blind spots represent a problem for a system with only one such camera.
Thus, the main problem for the AI which shall detect persons is how exactly to distinguish between a cup of hot liquid which is close to the camera and a person which is far behind? Since the shapes are not perfectly sharp, there might be used a human body template to look for persons in such images, but another question arises: how do you identify between two persons that are on the same line facing the camera? The image will show a bunch of heat and the AI may identify one person, but it is impossible to find the second person just by using that camera. So, blind spots represent a problem for a system with only one such camera.
Line 494: Line 495:
== Comparing the systems ==
== Comparing the systems ==


Of all the localization systems discussed above, one should be chosen to implement the system. The systems have all their pros and cons, the best system should be chosen. To choose the best system, the requirements should be addressed to make sure the system which can fulfill the most requirements is chosen. Below is a table with all the systems discussed, their downsides and advantages.  
Of all the localization systems discussed above, one should be chosen to implement the system. The systems have all their pros and cons, the best system should be chosen. To choose the best system, the requirements should be addressed to make sure the system which can fulfill the most requirements is chosen. In Table 9 are  all the systems adressed, with their downsides and advantages.  


{| border="1"
{| border="1"
Line 575: Line 576:
== Conclusion ==
== Conclusion ==


As described previously, the system will enhance safety for children and convenience for the parents. A system that is unreliable in 5% of the runs is not going to be agreed on by the parents. A system that fails approximately once in 20 runs is not safe for the kid and is therefore not worth being implemented, especially when the production costs are pretty high.  
As described previously, the system will enhance safety for children and convenience for the parents. A system that is unreliable in 5% of the runs is not going to be agreed on by the parents. A system that fails approximately once in 20 runs is not safe for the child and is therefore not worth being implemented, especially when the production costs are pretty high.  


In addition, the system must be able to track multiple people and differentiate among children and adults. Among all the possible detection solution described above, it can be concluded that the best options are IR sensors and thermal imaging. The IR sensors have a few drawback that the detection of the subjects might be impossible to be done, which finally leads to frustration for the parents and safety risks for their children. This method of detecting shall be able to also detect visitors and treat them like a member of the family: getting access whenever is needed and accepted by the rules of the system. The limited visibility of an IR sensor requires either multiple sensors to be used or a gimbal installed for getting an overall image of the security of the subjects. These small inefficiencies leave space for the second technology, the thermal imaging. The key points of this technology are the easy detection and differentiation between different people (all it’s drawback could be solved by installing multiple types of such cameras or few extra sensors that can provide the needed data for tracking). Compared with cameras (another possibilty to the given extent), the thermal imaging is not invading the privacy, the code is cheaper to produce and a bit more efficient compared with normal vision cameras. The accuracy is around 95% which is acceptable, but this number can grow when distances of the target can be measured so the system exactly knows where in the hallway the target is located.
In addition, the system must be able to track multiple people and differentiate among children and adults. Among all the possible detection solution described above, it can be concluded that the best options are IR sensors and thermal imaging. The IR sensors have a few drawbacks that the detection of the subjects might be impossible to be done, which finally leads to frustration for the parents and safety risks for their children. This method of detecting shall be able to also detect visitors and treat them like a member of the family: getting access whenever is needed and accepted by the rules of the system. The limited visibility of an IR sensor requires either multiple sensors to be used or a gimbal installed for getting an overall image of the security of the subjects. These small inefficiencies leave space for the second technology, the thermal imaging. The key points of this technology are the easy detection and differentiation between different people (all it’s drawback could be solved by installing multiple types of such cameras or few extra sensors that can provide the needed data for tracking). Compared with cameras (another possibilty to the given extent), the thermal imaging is not invading the privacy, the code is cheaper to produce and a bit more efficient compared with normal vision cameras. The accuracy is around 95% which is acceptable, but this number can grow when distances of the target can be measured so the system exactly knows where in the hallway the target is located.


Given the context of the SMART house application that shall provide automatic protection for subjects, thermal cameras are going to be elaborated further on. The design of the scenarios and the needed hardware will be adapted to the chosen detection system and the software implementation shows a possible way of detecting multiple people in the same image.
Given the context of the SMART house application that shall provide automatic protection for subjects, thermal cameras are going to be elaborated further on. The design of the scenarios and the needed hardware will be adapted to the chosen detection system and the software implementation shows a possible way of detecting multiple people in the same image.
Line 655: Line 656:
The camera needs to satisfy certain needs:
The camera needs to satisfy certain needs:
   
   
* has to be mountable on the ceiling (2.5 meters high)
* Has to be mountable on the ceiling (2.5 meters high)
* Needs to have a clear view of the entire hallway
* Needs to have a clear view of the entire hallway
* Needs an FOV of at least 90°
* Needs an FOV of at least 90°
Line 856: Line 857:
|}
|}


= Further improvements =
= Further Improvements =


== Hardware ==
== Hardware ==
Line 877: Line 878:
Since the current person detection is done by finding few black lines of vertical pixels between objects, any heat source that fits in the dimensions requirements can be misinterpreted as being a human. This ensures that the subject is connected with all his/her parts(even if they are few pixels apart from the main body, hands, legs and other small parts of the human body are considered as belonging to that person. The problem could be used by applying a modified BFS which shall test about 5 pixels in all directions, but it would be heavily inefficient since the complexity for a BFS with a huge number of possible neighbors increases exponentially.
Since the current person detection is done by finding few black lines of vertical pixels between objects, any heat source that fits in the dimensions requirements can be misinterpreted as being a human. This ensures that the subject is connected with all his/her parts(even if they are few pixels apart from the main body, hands, legs and other small parts of the human body are considered as belonging to that person. The problem could be used by applying a modified BFS which shall test about 5 pixels in all directions, but it would be heavily inefficient since the complexity for a BFS with a huge number of possible neighbors increases exponentially.


= Expansion towards other rooms =
= Expansion Towards Other Rooms =
   
   
The underlying idea of this project is the SMART house. The goal is to take the idea of a SMART house and research how to improve the child safety of it. It was found that the most child related accidents happen inside the house due to falling. Therefore the possibility to implement the SMART house idea to have a child safe stair is looked at. The next step is to see how this child safe stair can be implemented in the rest of the SMART house idea. Since a house contains more rooms that consist of potential danger for a child, for example the kitchen and the bathroom, there is looked at how the child safe stair system can be converted to other rooms.
The underlying idea of this project is the SMART house. The goal is to take the idea of a SMART house and research how to improve the child safety of it. It was found that the most child related accidents happen inside the house due to falling. Therefore the possibility to implement the SMART house idea to have a child safe stair is looked at. The next step is to see how this child safe stair can be implemented in the rest of the SMART house idea. Since a house contains more rooms that consist of potential danger for a child, for example the kitchen and the bathroom, there is looked at how the child safe stair system can be converted to other rooms.
Line 904: Line 905:


All the aspects discussed in the beginning how enterprise can benefit from a child safe SMART house system also counts for this intelligent subsystem. There will be increased products sales because there will be new products to sell and the pay-as-you-go house service can be implemented for the staircase. Also research still have to be performed because the concept is not fully developed yet. In the section ‘Further Improvements’ there are some problems stated which the system was not able to solve. To solve these problems research is necessary and the outcome of this research can be profitable for the enterprise.
All the aspects discussed in the beginning how enterprise can benefit from a child safe SMART house system also counts for this intelligent subsystem. There will be increased products sales because there will be new products to sell and the pay-as-you-go house service can be implemented for the staircase. Also research still have to be performed because the concept is not fully developed yet. In the section ‘Further Improvements’ there are some problems stated which the system was not able to solve. To solve these problems research is necessary and the outcome of this research can be profitable for the enterprise.
= Questionnaire =
For this questionnaire to work, the person that will be questioned first has to be enlightened about our project. So tell them about the idea of a SMART Home, that our idea is to implement a part that helps parents with the safety of their children and that we first start with making the area where the most accidents happen (the staircase) is being looked at first.
After explaining the basics of our project, the interview can be started. At first, some basic questions should be asked:
* What’s a number of children and the ages of their children?
* Do/did you own a stair gate?
After these basic questions, questions can be asked about our main subject:
*How would you feel about a SMART Home that helps you with the protection of your child?
*What home-related accidents did happen the most to your children?
*Do you think that making the house safer, by using locks, stair gates and shutting off the hot water influences the child in a bad way? For example, the child would not learn danger from hurting itself because it was always protected from it.
* Where did most of the home-related accidents happen?
* What inconveniences in keeping your child safe should the system be able to solve or help with?
After that, some more in depth questions about the stairs itself can be asked:
* At what age were your children capable of getting up and down the stairs on their own without endangering themselves?
* Did you experience any inconvenience of the stair gate, like opening and closing it all the time?
* Do you consider a stair gate as a good safety option?
* Do you consider an automated stair gate as something you would buy? Even if it has a higher price than a normal stair gate?
* What would be the best operating mechanism of the gate for you? Always closed but opens when you are around it, or always open but closes when the child is around it?
* At what age do children really listen to and follow the “commands” you give them? For example, move away from the stairs.
* Should the gate take into account that both you and your child can be in the danger zone? So should it do something different from the situations in which only a parent or a child is in the danger zone?
* If it should, how could the gate take that into account?


= Conclusion and Evaluation =
= Conclusion and Evaluation =

Latest revision as of 10:38, 26 June 2017

Group Members

  • Sjoerd van Helden 0893960
  • Stijn Middelhuis 0947014
  • Roy Niemark 0956824
  • Andrei Pintilie 0980402
  • Dennis Struver 0955477

Project Definition

In this chapter the project is defined by elaborating the subject, goal, approach, objectives, deliverables and the project planning.

Subject

The safety of a child. For a parent, this is all that matters. But keeping their child safe all time is a very time-consuming job. Having to constantly watch the child and making sure the activity the child is performing is safe costs a lot of time and effort. The child's home is of importance when analyzing child injuries. It is the environment in which young children grow up in and achieve developmental milestones. A child’s injury risk within the home is a joint interaction between the parents, the child and the environment. Both unintentional and intentional injuries are of importance as the majority of infant/child homicides occur in the home. Most of the accidents under children take place at home. The most severe injuries are associated with heat-related accidents and fall from high places, so even if the most accidents were encountered in the living/dining room, the most serious ones happen in the kitchen and on the stairs. Regarding the falls, around 10 children die each year by this accident. But, mostly, children that fall from stairs or high places encounter trauma and possible some visible post-accident problems.

Technology has been making a lot of progress over the years and this will keep on going. Since there are a lot of possibilities nowadays, more ideas can be realized. One of the new technologies that are growing fast is a SMART House system[1]. A SMART House is becoming more well-known and will extend more as we go into the future. The concept that has been conceived is an extension to the concept of a SMART Home. This SMART home will not be focusing on solely on luxury or caregiving, but on enhancing the safety of children. This system will focus on preventing certain dangerous situations and reduce the accidents that occur at home. The concept will be installable in every house with the implementations that are relevant. ‘The house’ will then provide certain actions with the equipment that it has available to keep the children safer. The system needs to be able to detect where a person is in the house or in a room, but also recognize if it is a parent or a child. When it is able to do that, it also needs to be able to communicate with its other components that can prevent danger in the house, and take action when the child is approaching such a dangerous situation. When it is able to distinguish parents and children, the parents will not experience any obstruction from the system. Also, it does not need to take the same actions when the child is with a parent because this situation is a lot less dangerous.

The concept can be used for children of an age between 1 and 9 years old. In this age-category, children are most likely to have serious accidents at home. Also, children at that age are able to move around the house and interact with objects by itself but are not able to stay at home by themselves. The accidents and injuries that are going to be focused on reducing are injuries due to poisoning, falls and thermal injuries. The concept of this SMART Home will keep the child away from dangerous situations and/or objects, and when it is not able to take appropriate action it will alert the parents. For example, when the system detects that the child is near a sink with hot water, it shuts down the hot water so the child can not burn itself.

Goal

Since today's technology goes pretty far already, there is a lot possible considering a SMART Home. Yet when this needs to be adapted and be a perfect fit for your child's safety, some factors and aspects need to be researched and designed for this purpose. Our goal is to reduce accidents and injuries of children in their home by extending the concept of the SMART Home with respect to child safety while keeping the technological, ethical and financial aspects in mind. In this project, a specific part of a home will be addressed and a specific accident category will be targeted. To extend the concept of the SMART Home we will research and design the system needed for the part that is going to be addressed, evaluate this in detail and look at the concept and how this can be applied to other parts of the house. How this will be achieved will be explained in the Approach and Objectives.

Approach

To reach our goal it is necessary to gather specific and discrete information about frequent occurring accidents and injuries of children. Then such a SMART Home system with its focus on the safety of the child will be researched and the possibilities and the techniques will be exploited. Therefore, detection hardware systems that satisfy the requirements of this system will be analyzed and researched and the best option for this concept will be recommended. The requirements for this system and what the possibilities are will be elaborated further on in this project. Since the detection is the most important part of this system this needs to be viable in order to let the system work and will get the most attention. In this project, a part of such a SMART Home system will be addressed and elaborated. The focus will lie on one accident category and in the end of the project a specific part of the system will be designed and evaluated in detail. Finally, the designed part of the SMART Home system will be applied to more parts of the house. With example situations and simulations, the designed system for the specific part and other parts of the house will be evaluated. In this project, it is important to evaluate the impact of the system, to stay close to the users and user needs, to address USE-aspects and to keep in mind the cost and benefit. Thus when designing the system the users, cost, and benefit will receive a lot of attention and the final stage of the project will be dedicated to the evaluation with respect to impact, USE-aspect, and overall benefit.

Objectives

Following the approach and considering the goal of this project the following objectives can be derived:

O1: Gather information about frequent child accidents and injuries in the home.

To be able to implement a system that enhances the safety of children, it needs to be clear where the most accidents happen, what kind of accidents actually happen and how severe the injuries are. Then different situations can be considered and different safety issues can be targeted.

O2: Research the potential dangers, and potential prevention.

When the kind of accidents that happen frequently are known, we want to know how these can happen and how this can be prevented. When the potential dangers are known, it is important that the system is able to recognize these dangers or to know how it can reduce them. The dangers that have to be prevented need to be implemented into the system, and the action that belongs to preventing the danger also has to be implemented.

O3: Find the best option for detection and recognition.

To be able to distinguish an adult from a child, the system has to be able to recognize different subjects. The system will need to operate mostly when a child is alone or not supervised.

O4: Define and work out a specific accident category and situation in the house.

We want to narrow down the concept due to the time window that is available for this project. We will choose a specific situation and target a specific accident category that is of higher importance but also that will use a concept that is suitable for different situations as well.

O5: Design and evaluate the part of the system for the specifications determined in O4.

Consider all the important factors, such as user, ethics, and cost, to design a concept for the specific situation that will reduce a certain risk of an accident category. Take into account safety margins and user needs, but also possibilities for extensions.

O6: Simulate different situations to show usability, operating and impact of the concept.

When a concept is designed we will use simulations to visualize how it works and what it can do. Also, this will show extensions and possibilities.

Deliverables

In this project, the following things will be delivered at the end.

  • Documentation of the research and literature study
  • Preliminary design: Stairs (drawings)
  • Scenario descriptions of the designed system for the stairs (visualization).
    • Gate
    • Sensors
    • Stairs
    • Child
  • Evaluation of the system with respect to the impact, USE-aspects, and cost-benefit

Project planning

For a project, is important to have a good project planning. Below the planning for this project is given. The week planning and a role distribution are made to keep track of the progress and provide a guideline while working on this project. The planning follows the process and sets several milestones that are important to achieve. The planning is made at the begin of the project and has some room for adjustments, if necessary. In week 4 we made some changes to the project definition and narrowed down the subject. This had a significant impact on our planning and thus the planning is adjusted with the changed definition, to make sure the milestones are achieved properly with the time that is left.

During every week, the wiki should be updated with the progress made up until that point. The last week is dedicated to preparing for the final presentation and finalizing the wiki. The week planning below contains some more specific detail to the different steps that need to be taken in this project.

The general approach of this project consists of the following milestones (for more detail see the planning itself):

  1. Research background, state-of-the-art and similar existing systems
  2. Research detection and recognition technologies with the respect to our subject
  3. Draw conclusions and recommendations for the detection possibilities
  4. Designing a part of the system focused on a specific situation
  5. Evaluate the designed system with simulations and work out possible extensions or adjustments
  6. Evaluate the cost, benefit, and impact of SMART Home system and the detection possibilities


Weekplanning

Week 1

  • Determine the subject
    • Formulate the problem
    • Create idea’s for a concept
    • Objectives
    • Involved users
  • Research about background, state-of-the-art and similar existing systems
  • Create planning and presentation

Week 2

  • Continuing the research
    • Children and accidents
    • Existing SMART Homes and its collaboration with safety measures
    • Typical house environments for children
    • State-of-the-art technology that could be implemented or used.
    • Existing systems made for safety of children
    • User benefit
  • Determine important and critical points of interest
  • Look into the subject from a USE perspective and determine relevant USE aspects

Week 3

  • Finish the background research (Milestone 1)
  • Start research about the detection possibilities
    • Existing technologies
    • Requirements and options for this project
    • How and what can be accomplished in our project with which technology
    • State-of-the-art options
  • Conceptualize the subject to a specific and detailed design question (Milestone 2)
    • Composition of the room
    • Components (technological) which can be used
    • Elaboration of the design requirements

Adjusted planning for weeks 4 up until 9:

At the beginning of week 4, we decided to narrow down our project. Instead of looking at all possible dangers for children in a home and how to tackle them, we now only focus on one of the biggest causes of both fatal and non-fatal injuries, falls from the stairs. This means that we do not have to consider the whole house, but only the stairways and the halls towards the stairways. Therefore we have to change our planning. Here is the new planning for the following weeks:

Week 4

  • Narrow down the subject and elaborate on the choices
    • Assumptions made
    • Determine the deliverables
    • What is taken into account and what is not?
    • Why are they taken into account or not?
  • Research the possibilities of localization
    • Proximity detection between child and certain points
    • Badges/bracelets
    • Ultrasound
    • Infrared sensors
  • Research the possibilities of person recognition/detection
    • Difference between a child and an adult

Week 5

  • Research the necessary information for design a system
    • State-of-the-art technology (for staircase gates)
    • Movement speed of the children
    • Closing speed of such a gate
    • Safety settings for the system in combination with the localization (safety margins to be sure the gate is closed in time)
    • System settings (specified to age and preference of the parents)
  • Designing the system following the design question
    • Which localization and recognition/detection will be used?
    • What technical hardware will be implemented?
    • How will this be programmed and setup?
    • How will the system work?

Week 6

    • Finish the design for the specified purpose
    • Describe several scenarios
    • Pros and cons of the designed system
    • Possible extensions and settings
  • Evaluated the design for the specified purpose
    • What can/will it do?
    • Possibilities and usability
    • Cost of this part of the system
  • Application of this part in a SMART home system
    • Extension to a bigger system
    • Application of this system in other parts of the house

Week 7

  • Work out the application in a whole house
    • Different options for localization/detection
    • System settings and extensions
  • Evaluate the designed system
    • Benefit for the users
    • Impact from the perspective of USE

Week 8

  • Accomplish recommendations and conclusion for designing such a system
  • Elaborate on the different detection/recognition and localization possibilities and their pros and cons
  • Elaborate and conclude the evaluation and impact of such a system

Week 9

  • Finish, prepare and give the presentation
  • Finish the wiki
    • Reorganize if necessary
    • Check the progress
    • Complete the final wiki page

USE-aspects SMART House

Users

The primary users of the system are the parents of the children and the children themselves. The parents are the one that will buy the system and have the system installed into their house. The parents will expect the system to help them protect their children and keep them safe. The children are the ones where the system is designed for and are therefore also primary users. However, the children will barely know that the system is there since they will not see the technology behind it. The secondary users of the system are older children, nannies and other people who visit the house. The secondary users will know that the system is there and will sometimes notice its actions, but will not be affected by it most of the time. When a secondary user is in the room with a child that is protected by the house, the actions of the system will be slightly different since the child is under supervision at that moment, even if that supervision is an adult visitor of the house.. The tertiary users are the technicians, mechanics and software engineers who make and implement the system. The technicians have to make sure that the system is easily installed, removed and also calibrated if necessary. They will also be the ones that conduct maintenance or come with updates.

Society

Parenting can be stressful and there can be multiple reasons for that. Among those reasons are: time demands, relationship demands (related to time demands), protective instinct/uncertainty and a lack of alone time. Also the stress level for single parents is often higher when a single mother have to take care of their young children. [2][3]

Stress is bad for the parents itself but it also affects the child negatively. Parental stress can lead to mental health problems (depression, anxiety, internalizing behavior) among children already at a young age and it negatively affects their externalizing behavior. Externalizing behavior is usually associated with multiple disorders like Antisocial Personality Disorder, Oppositional defiant disorder, pyromania among others. It has also been shown that parental fatigue can have a negative effect on a child but also affects parental practices .[4] [5] [6] [7] [8]

The problems that result from parental stress and fatigue are detrimental for society. Children with mental health issues will need to get treated for that which cost money and time and it might never be fully healed. For example, the risk of recurrence after a first major depressive episode is 50% and increases with subsequent episodes (Post, 1992, Kupfer et al., 1996, American Psychiatric Association, 2000). Children with mental health problems will perform worse in school and other places than healthy children. If their mental health problems are never healed it will affect their adult lives as well. Parental fatigue might lead to bad parenting which also is not desirable.

This means that society benefits from a solution to parental stress and fatigue which this SMART home system provides. The child safe SMART home can tackle the reasons mentioned in the beginning and help reduce parental fatigue. But also the child safe system can reduce the accidents which happen among young children. This reducement in the amount of injuries could positively affect the health statistics of society. Later on there will be a recapitulation about if this is achieved with the designed system.

However, a SMART home that protects a child from any kind of harm within the house could also be negative for the child’s development and mental health. Multiple studies have shown that overprotective parenting affects the child’s mental health, anxiety disorder is the most common one. This would not be beneficial to society because this would impair a child’s productivity and quality of life. More recently the Dutch institution VeiligheidNL argued that parents should allow their children to take on more risks because that would be beneficial to their development. The potential benefits that a SMART home would have for society would be offset by these problems. Therefore the child safe system should be implemented wisely in a SMART house system to avoid these problems.[9][10] [11][12][13][14][15][16]

Enterprise

Being able to stay in touch with your customers is the best outcome that a salesman can achieve. The business model for SMART houses is just at its beginning, which means that there is plenty of space for new developments and ideas that can create a strong bond between the users and the merchants. Even if there are few sectors that might suffer because of this newly created area, most of the actors in the project are going to benefit. As main enterprise actors can be included the retailers, companies that provide technology and the safety companies. In the vision of this project, a safe SMART house is intended to keep the children safe from most dangers that can occur with help of SMART systems.

But how exactly will enterprise profit by this new area of interests? First, it is necessary to be specified that the entrepreneurs will take a significant role in the safe SMART house area. They are going to provide the necessary technology and safety regulations. So, the next three points represent the main interests of the enterprise:[17]

  • Increased products sales
  • Pay-as-you-go house services
  • Providable research outcomes

The first point is stated as Increased hardware sales. A safe SMART house contains more than just ordinary hardware that has to be created. It needs SMART products which have been tested and do not represent a problem for most types of users. Also these products have to implementable in a SMART house. A product that shall be able to connect to the house facilities and integrate properly. As you can speculate, this means a lot of revenue from selling these items. A second point is represented by the pay-as-you-go house services. Besides the payment for the SMART technology that might occur, users shall be able to pay for their health and safety, which means that the retailers could improve the software of a machine to increase the safety level on the amount of money you are willing to pay. Also, new devices could be added later as you pay when you think you need them. The last point is represented by the research which still have to be done. Full functioning SMART houses are still in development and therefore research is necessary. This counts for the convenience part of a SMART house, but also for the child safety which is treated in this project. Research can result in new or better products which maybe can also be used outside the SMART house application. Enterprise benefits from this by putting these new products on the market.

A SMART home would enable retailers to develop a close relationship with their clients through the SMART devices which are placed in the home. It means that in the case of a problem that occurs, retailers will be able to find a solution or to recommend new products directly to the customers. When maintenance is needed this can also be done quickly by making directly an appointment. The greatest achievement is represented by the possibility to seduce young people, “tech savvy consumers” and the fast possibility of adds presentation and close connection to the users.[18]

The enterprise point of view of this project would not be much different than a normal SMART house. The addition of making it child safe will involve some more actors like people that provide hardware tools for safety and software developers that need to develop more than just a self-aware house that can notify the user. It should also react and protect the children. As an entrepreneur, everything related to this idea is reduced to money. On the other side of these new possibilities offered by this technology is the security aspect, both in physical and software ways. In the software ways, SMART houses are going to be the target of hackers, which can use the technology designed to keep you safe in wrong ways. That’s why the entrepreneurs have to invest a lot of money in secure software. Besides the financial costs, the missing security in software can turn the population against the usage of such a technology.[19]

Accident Statistics

Reason for research

Nowadays, there is a big factor that influences the lives of children all around the globe which is represented by the dangers in their own environment, in their own homes. According to RoSPA[20] more than £275 million a year is spent for these types of accidents and most of the accidents among children take place at home[21]. Children below nine are more exposed than the older kids, mostly because of unconscious acts that they do not percept.

As presented in Figure 1, the most accidents among children in the age of 1-9 happen at home. For this project it is necessary to know which accidents happen and where they happen inside the home. This way the system can be implemented at the right locations where it is needed the most within a house. Research has been done to fatal and non-fatal injuries of children within the age group due to accidents in the home environment. But the impact of injuries is much greater, with millions of hospitalizations and emergency care visits. Although injuries are a leading cause of the burden of disease and seriously drain health and societal resources, they have not been a high-priority area for action in most countries until recently.

Figure 1: Accidental injuries according to age and place of accident.


Table 1: Fatal unintentional injuries and their location
Category n %
Transportation-related injuries 44.830 49
All other injuries
Home 18.018 20
Other 11.762 13
Unknown 14.596 16
Blank 1.622 2

USA and EU

For the research the focus lies on the USA and Europe. The injuries can be divided into different categories. The unintentional home injury fatal injuries and unintentional non-fatal home injuries will be elaborated so that a good conclusion can be drawn, which will be done at the end of this paragraph.

Unintentional Fatal Home Injuries

Number of Accidents

From 1992 to 1999, there was an average of 146.970 injury-related deaths annually in the United States, with an average annual injury death rate of 54.90 per 100.000 persons. This is a total of all age groups and all possible injuries, in and outside the house. In Table 1, the location of the fatal unintentional injuries is listed.

Of the injuries with a known location, an average of 18.048 unintentional injury deaths occurred annually in the home environment. This represents an annual rate of 6.83 deaths per 100.000 persons. In Table 2 is the fatal accidents per age and per gender listed. Children in the age of 1 to 9 account for 7,5% of these fatal injuries at home in the USA.

Table 2: Percentage of fatal accidents for each age group
Table 3: Location of fatal injuries by age, average annual number and percent, 16 European countries, 2002-2004

In the EU unintentional injuries are the leading cause of death among children aged 5-19 years. In 2004, 42 000 children and adolescents aged 0-19 years died from unintentional injuries in the WHO European Region.

The home is of particular importance when analyzing child injuries as it is the environment in which young children grow up in and achieve developmental milestones by interacting with their physical surroundings. In the United Kingdom alone, 75 children under 15 years of age died due to home injuries in 1 year, ∼25% of all child injury deaths. A child’s injury risk within the home is a joint interaction between the caregiver, the child, and the home environment.

In different countries from the East and West of Europe with variations in income levels. 60% of injuries to children under 1 year of age occurred in the home environment, compared to 11% in transport. In contrast to fatal transport injuries, which increased as age increased, Table 3 shows the fatal home injuries as highest in children under 5 years of age both in numbers and proportions of the total and then sharply decreasing. For children between 0-9 years the fatal injuries at home account for 38% of the total fatal injuries. [22]

Causes

For these accidents, there are several major causes. In Figure 2, the major causes of injury fatalities by a range of age groups in the USA are given. Apparently, fires and burns, inhalation and suffocation and drowning were the leading causes of unintentional home injury deaths among children aged from 0 to 15 years old. Fire or burn deaths were the leading cause, with 43.7%, for children from 1 to 9 years old and drowning the second leading cause, with 29.3%. Nearly all fire/burn injury deaths among children aged below 15 were the result of residential fires. The majority of drowning deaths among infants occurred in bathtubs.

Figure 2: Number of fatal accidents for each age group
Table 4: Number of home-related accidents for 2000-2008

In Table 4 it is shown that from 2000 to 2008, there was an annual average of 30.569 unintentional injury deaths occurring in the home environment in the USA. This number is almost twice as high as in 1999 where the annual unintentional home injury deaths were just more than 18.000. In Figure 2 it becomes clear that the most unintentional injury deaths are caused by poisoning, fall, fire/burn and choking/suffocation. This does show that these causes are also the biggest issues from the year 2000 to 2008. In Table 5 the age-adjusted rate of unintentional home injury deaths for 2000-2008 is shown.

Table 5: Age-adjusted rate of unintentional home injury death’s
Table 6: Home injury deaths in 16 European countries, annual number, percentage, age 0-14, all causes, 2002-2004

The 3 major causes of unintentional home injury death of the age group 0 to 9 are still suffocation, drowning and fire/burn, which is the same result as the research from 1992 to 1999. It is clear that the fatalities inside a house for children stay about the same looking at the results from 1992 to 2008. Therefore it is assumed that these causes are nowadays still present and that the most fatal injuries are caused by suffocation, drowning and fire or burns.

The leading mechanisms of death from unintentional injury in children are road traffic crashes, drowning, poisoning, thermal injuries and falls. The top four causes of unintentional fatal home injuries in children 0–14 years in all countries aggregated were drowning/submersion, fire/flames, poisoning and falls, see Table 6. These causes accounted for almost 75% of all home injury deaths[23]

Unintentional non-fatal home injuries

The non-fatal injury rate for children younger than 9 years old In the USA was 28.054 per 100.000 in 2000-2006. For this age group falls accounted for the largest amount of injuries. In Table 7 the top five leading causes of unintentional injuries in the USA are displayed. Getting struck by or against an object was the second largest cause of injuries. For children aged 1-4 cuts or pierce injuries account for 4% of the total amount of injuries (compared to 43% for falls) and for children aged 5-9 cuts or pierce injuries account for 7% of the total amount (compared to 37% for falls). When children grow older outside related accidents (transportation) become more important.

Table 7: Top Five Leading Causes of Unintentional Injury Deaths and Nonfatal Injuries among Children 0 to 19 Years, by Age Group, United States, 2000–2006
Table 8: National estimates of number of nonfatal, unintentional injuries, by location, United States, 1998–1999

Falls from furniture and child care products are the most predominant ones for children younger than 1-year-old. For the older age groups injury by falls still predominantly are indoors(falling of stairs, the bed and tripping over objects like toys). For children younger than 5 years old falling out of the window is significant compared to the other age groups. Burns are most often happen in the kitchen or in the vicinity of hot water(bathtub), but hot beverages and food can also cause burn injuries. Especially children younger than 2 years old are at risk to be burned.

A study conducted in the late 1990’s also showed an alarming number of accidents that happen in the home environment. The NHAMCS data indicate that nearly 9.8 million emergency department and 1.4 million outpatient hospital visits were made in 1999 for nonfatal, unintentional injuries that took place in a home environment as shown in Table 8. Likewise, data obtained in the NHIS include 12.922.220 nonfatal unintentional home injuries, excluding poisonings, requiring some form of medical advice. NHIS data also show that 750.000 persons aged 5 years were reported as missing at least 1 day of school, as a result of an unintentional home injury.

Each data set identified falls as the most common mechanism of injury by far, accounting for 36.2% to 45.7% of the injuries or visits to healthcare providers for nonfatal unintentional home injury. The national estimates of the numbers and rate of having lost at least 1 day from work or school due to falls were 214.5044 and 757.044. Almost 4 million emergency department visits and 4.2 million office-based physician visits were made because of a fall in 1999. The second most common mechanism of injury varied according to the data source. For visits recorded in the NAMCS, NHAMCS-OPD, and NHAMCS-ED data sets, being struck by or against an object was the second most common mechanism indicated, with visit rates per 100.000 at 439, 63, and 591, respectively. In contrast, the NHIS dataset identified cuts and piercing injuries as the second most common mechanism; the visit rate was 649 per 100.000.

According to the data set people older than 65 are at the highest risk of getting injured followed by children younger than 14 years old as shown in Figure 3. This shows that young children are at risk of unintentionally injuring themselves via an accident at home. Among these children falls, cut/pierce and struck by/against are the main causes for an injury which is consistent with the data from early 2000.

Figure 3: National estimates of the rate of non-fatal unintentional home injuries, or injury visits, by age group, United States, 1998–1999
Figure 4: a) Main places of occurrence of non-fatal child injuries by age group. b) Mechanisms of home injuries in children under 5 years of age.

In the EU non-fatal injuries at home account for about 45% of the non-fatal injuries, as shown in Figure 4. For children under 5 years, 59% of these injuries are caused by falls from heights or other falls. As age increases, the home injuries decreases. For children from 5 to 9, 38% of the non-fatal injuries happen at home. [24]


Stairs

In the USA, 931.886 children aged less than 5 years were treated for stair-related injuries from 1999 through 2008, averaging 93.189 injuries per year and 46.5 injuries per 10.000 population annually in the USA. This shows that falling of stairs is a major health risk for children. [25]

In the EU the top four causes of unintentional fatal home injuries for children between 0-14 years were drowning/submersion, fire/flames, poisoning and falls. These causes accounted for almost 75% of all home injury deaths in Europe. Of this 75%, about 15% of the fatal home injuries are because of falls.

In comparison to fatal injuries, non-fatal injuries at home account for about 45% of the non-fatal injuries in Europe and for about 34-46% in the USA. In Europe, for children under 5 years, 59% of these injuries are caused by falls from heights or other falls. As age increases, the home injuries decreases. For children from 5 to 9, 38% of the non-fatal injuries happen at home. In the USA, for children between 0-9 years falls accounted for the largest amount of injuries. Falls from furniture and child care products are the most predominant ones for children younger than 1-year-old. For the older age groups injury by falls still predominantly are indoors (falling of stairs, the bed and tripping over objects like toys). For children younger than 5 years old falling out of the window is significant compared to the other age groups.

Conclusion

In USA and the EU a significant amount of injuries, fatal or non-fatal happen among children. Falls in the home environment caused a majority of these accidents as shown in the research above and reducing these accidents can make a real difference. Also is seen that a big number of these falls is due to accidents which happen at the stairs. This project will therefore focus on making the stairs of a SMART home safer, by giving the SMART home system several implementations to take action and reduce the amount of accidents that happen. In the end, the first design for a part of such a SMART home system as being addressed in this project will be for the safety enhancement around the stairs.

State-of-the-art Childproof Equipment

Unreliability of Childproof equipment

Nowadays, childproofing a house is using simple measures like locks and blockers to increase the safety of a house for a child. A child is not able to open these locks and dangerous situations are in this way avoided. For example, plug covers or plastic locks for a cabinet or bottles are used to childproof a house, in short, it’s denying physical access to a potential danger. Most of this child proof equipment is designed in such a way that when it is installed, grown-ups can still access these dangers but children do not. This means that childproofing is giving up some comfort in favor of child safety. For example, a magnetic lock for a cabinet cannot be opened without the use of the magnet key which can be inconvenient if a grown-up does not have that key at a specific moment and needs to go get it from a different room. [26] [27]

Giving up comfort is not the only problem with these simple measures, these safety measures do not always guarantee safety for the child. This is because a child becomes stronger and more clever when they become older and also because these safety measures are designed in such a way that adults and seniors are still able to open them. Take for example safety caps that are used on medication and cleaning products. A 2-year-old has a 2% of opening such a safety cap and if the child has observed an adult opening such a cap, then this chance rises to 8%. So for very young children it can be stated that these measures work properly. For a 7-year-old the chance to open it is already 60% and rises to 74% if it has observed an adult. The chances of opening such a cap increase with age and it shows that they are not reliable. Especially children who get the chance to watch their (grand)parents opening such a safety cap are able to open it later. A child only needs a couple of seconds to a minute to open a cap which means that parents have to keep an eye out constantly. [28] [29] [30] [31]

One of the major reasons that children are able to open these caps is because seniors need to be able to open them as well. Seniors often need medication and if the safety cap is too difficult to open, this can cause problems for the seniors' health. A 60+ year-old can generate about twice the torque than a 3-5-year-old. However, 8-12-year-olds can already generate about 80% of the torque that a 60+ year can generate. This means that caps that can be opened by seniors, which are most of them, can also be opened by older children but also by above average younger children when it comes to strength or intelligence. Another reason why these caps can be opened is that of comfort. Adults who open such a cap will put stress on their hand's joints which cause discomfort. If the caps are too hard to open then people will not buy them because of this discomfort. Also notable is that a 3-5-year-old is able to generate more torque using a 3-fingered grip than any adult using a 2-fingered grip. The 3-fingered grip is the most common and the more natural grip to use to open such a cap which shows that a young child is not weak by any means.[32] [33]

These examples show that children learn from their parents to open childproof safety measures which make these safety measures less reliable. Also, these safety measures are not made to be extremely difficult to open because adults need to be able to do that. The cleverness and persistence of children also should not be underestimated. For example 1 in 4 children in Ireland successfully opened the door of a moving car and about 76% was able to free themselves from their straps. The only solutions to combat such problems is for the parent to be very observed of their children, but of course, this is not always reasonable. Parent are humans and therefore will make mistakes or forget and that almost always results in problems with their child. [34]

Simple gate

Types

There are three basic types of Baby Gates: Pressure Fit Baby Gates, Hardware Mounted Baby Gates and Child Safety Gates for Wide or Irregular Areas:

Pressure Fit Baby Gates

In the past they were simply barriers that were wedged between two walls or in a doorway and you had to step over them in order to go through or remove them completely. Much has changed. Nowadays the gate itself stays across the opening and is held in place by the pressure that is usually created by extending threaded pressure pads to the wall or door jam, these gates have a door that can be opened and closed for convenience. Because this type has a gate within a gate it also has a threshold across the bottom when the gate is opened. Two things all pressure gates have in common is that they require two flat surfaces across from each other to be mounted against and they cannot be mounted on an angle. These gates generally are not designed to be mounted on staircases.

Hardware Mounted Baby Gates

These gates are sometimes also called stairway gates because they are the most appropriate type of child safety gate for a staircase. They are versatile and when mounted properly they are easy to open. Most can be removed easily from the mounting hardware if there are occasions when having a baby gate installed is not appropriate. These gates also come in different shapes, sizes, and colors. Some have the added ability to be able to be mounted on an angle if necessary. The ability of these gates to mount at different angles vary, the gate that can be mounted at the steepest angle is the angle mount safe way.

Child Safety Gates For Wide Or Irregular Openings

Wide or irregular openings usually take a little more time to plan but may actually end up being very simple to install. These situations require you to do a little creative thinking. [35]

Accidents

Baby gates are one of the most widely used home safety products to protect children from home hazards. The objective was to describe the epidemiology of baby gate and barrier-associated injuries among children. It was hypothesized that injuries experienced by children ages ≤2 years and those >2 years were significantly different as a result of differences in gate interactions. An estimated 37,673 children were treated in emergency departments for injuries associated with gates, yielding an average of 1794 cases annually. The incidence of gate-related injuries increased significantly from 3.9 per 100,000 children in 1990 to 12.5 per 100,000 children in 2010 (P < .001). Patients were primarily boys (61.0%) and were <2 years of age (60.4%). Patients <2 years of age were most often injured by falls down stairs (odds ratio 6.72; 95% confidence interval 6.32–7.16) after the collapse of the gate. Patients aged 2 to 6 were most often injured by contact with the gate (odds ratio 2.03; 95% confidence interval 1.95–2.12), resulting in open wounds (55.4%) and soft-tissue injuries (24.2%). Given the clear dichotomy between injury characteristics of patients aged <2 years and patients aged 2 to 6 years of age, as well as the prevalence of preventable injuries, greater efforts are needed to promote proper usage, ensure safety in product design, and increase awareness of age-related recommendations for use of gates. [36]

Design Requirements

For the design of this part of the system, first, the design requirements are set. These elaborate on what conditions the system should satisfy and which specifications should be met.

Stairs

Figure 5: Possible hallways.

To target the largest group of potential customers, the system should be able to operate at as much different house settings as possible. The staircase itself can differ per house and the system should be able to be implemented in every house. The design should be applicable for every type of stairs. For example U-shaped stairs, normal straight stairs or curved stairs.

Hallways

The hallways can also differ per house. To make this design albe to be used for most houses, the most common hallways settings need to be determined. The way the stairs are connected to the hallways determine important parameters for the safety. Some situations contain a higher risk than others due to the distance between the stairs and the closest room, or due to limited visibility. This design then should be applicable to hallways and stairs combinations that on the top side are shaped as L or U, or stairs that end in a cross- or T-intersection with the hallway. In Figure 5 below the most common possibilities are showed.

Gate

The system will prevent children from falling down the stairs and will enhance the safety of the child around the staircase. While fulfilling this task, the system should also not be inconvenient for the parents or other people in the house. It should not be an obstacle of some sort and still function properly. In the design, there should then be a gate on the top and bottom of the stairs. For convenience, the gate should move only in a plane direction. If the stairs are designed in such a way that a gate will not cover the entry of the staircase, then this should be solved with an extension of the gate but separately installable. The material of the gate should be sturdy and strong, but also not at harm for children. The edge which closes the gate to the wall should not be too hard and child safe and sharp edges should be avoided. The height of the gate should be equal to the standard that is set for these safety gates, this is about 0.7 to 1 meter in height. Since these gates should open and close automatically, it should satisfy a “certain minimum and maximum speed’’. So that it is safe and realizable, but also quick enough to be closed in time while not becomming an extra danger for the child.

Danger Zone

The system should have an implemented danger zone or reaction zone in which the gate should open for the parent and close for the child. This zone has a safety margin so that the gate is always closed when the child is in this zone. To notify the parents when the child is in this zone, LEDs should make this visual from all directions. This notification part should be implemented in the gate so that it does not have to be installed separately. This way the system can enhance safety and comunicate with the parents.

Recognition and detection

The system is there for the protection of the child(ren) in the house, but also to reduce the stress of the parents. Therefore the system should be able to accurately distinguish between adults and children. Besides recognizing the people in the house, it should also be able to detect where they are. It is of importance that the system is able to track the distance between its goal, in this case, the stairs, and its object of interest, in this case, the person (either an adult or a child). This detection and recognition sub system is key for this system to work, since it will take care of the convienience. If it does not work properly, the system fails his goal where it is designed for; enhance safety and bring more convienience for the parents.

Operating and working of the system

The system must be able to work with the information it gets from the detection and recognition software. The information that is acquired must be processed and converted into actions. This processing and converting must be done almost immediately. Therefore the system must be able to work together with multiple components at the same time.

Recognition and Detection Systems

For the system to work properly, a proper detection and localization system has to be chosen. The system has to know if there is a person in the defined space and to know its location to at when necessary. It is also necessary for the detection system to be able to differentiate between children and adults because the system has to act when a parent is near the gate. A few different detection systems have been investigated and compared to choose the best one. There is also looking to the set requirements to make sure the system satisfy these requirements.

State-of-the-art system

Infrared Detection

One way to locate an object is by making use of infrared (IR) sensors. Simple IR sensors are widely used nowadays in robotics and automation, process control, remote sensing, and safety and security systems. More specifically, they have been used in object and proximity detection, counting, distance and depth monitoring, floor sensing, position measurement and control, obstacle/collision avoidance, and map building. Since the great variety of applications of IR, IR sounds promising for the detection of a person and the differentiating between adults and children.

IR is also commonly used for face recognition systems, most of the time for safety reasons. Thermal face recognition deals with the face recognition system that takes the thermal heat of the face as an input. Thermal human face images are generated due to the thermal human body heat. Such a face recognition system would be also applicable for the differentiating a child from an adult. The generated thermal human face image of the observed person could be compared with an image of the child out of a database and in this way the system is able to differentiate the two. A downside of such a system is that the resolution of the thermal images is not very high. Also is it very plausible that the child will not always look straight into the camera, so different face position and also facial expressions should be covered by the system. These downsides together with the high cost of such high-end IR cameras make this application of IR not interesting for this application. [37]

Since the costs of such a system, to be able to cope with the state-of-the-art, have to be as low as possible, the localization system should not be too expensive. Therefore the applications of low-cost IR sensors is investigated. To still be able to differentiate objects, the differentiating techniques employed should be different from such operations performed on conventional images. The targets which have to be differentiated are not patterns in a 2D imaged, but rather objects in space, exhibiting depth and at different positions with respect to the sensing system. This would be a common situation which often will occur when the system is in use in a SMART home.

There are different techniques one can use to differentiate the geometry of an object with such low-cost IR sensors. Here, only the template based approach to differentiate only the geometry of the target object is discussed, since only the geometry is important for this application. The template-based approach is based on comparing the acquired IR intensity scans with previously stored templates acquired from targets with different properties. Hence, this approach relies on the distinctive natures of the IR intensity scans and requires the storage of a complete set of reference scans of interest. For targets made of the same material, but with different geometrical properties, the correct differentiation rate is 97 percent.

The geometry of children and adults is a major difference since a child is much smaller than an average adult. Therefore only a rough estimation of the size of the object is enough to differentiate a child from an adult. The geometry itself is therefore not important, one could for example approximate the persons geometry as a solid block or cylinder with specific dimensions of the length, width, and depth. To determine the full geometry of a person is also way too complex for such a low-cost sensor. Since it is possible to determine the size of an object, this technique would be suitable for this application. [38]

So far this system sounds very promising. The sensors used are of low cost, it is able to detect whether there is a person in the room and the differentiation of the geometry is of great accuracy. There are only a couple remarks to make.

The system determines the geometry by sensing the object, therefore the sensor should be mounted on a gimbal which can rotate the sensor in all directions. If the sensor is not able to rotate, it will not be able to fully sense the target and detection and differentiation becomes impossible. Since the system can only determine simple geometries, objects as for example furniture can cause false positives and negatives for this system. The system can track these objects and because the geometry of a cabinet can be the same as the simple modeled geometry of a child, the simplicity of this system can become a problem. The last problem is that it cost time for the system to fully observe the target since the full geometry has to be sensed. Moving targets, therefore, become a problem, because a child that moves will not be fully scanned by the system and therefore detection and differentiation become impossible. The simplicity of such low-cost IR sensing systems is preferable and promising, but due to the downsides of it as listed above, this way of using IR will not be suitable for this application.

WPIR

Figure 6: The functionality of a PIR sensor

Infrared can also be used in a different way. In the section above is discussed how IR sensors can be used to locate objects in an environment using the intensity of the reflected light. However, IR can also be used to locate objects by thermal mapping. There exist an indoor localization and monitoring system, which is based on a wireless and PIR (WPIR) sensor fusion system. The PIR sensor transforms incident IR radiation into an electrical signal. PIR detects changes in temperature coinciding with movement of a person or object in the environment. The output of the PIR sensor is in disorder for human movement detection. A human walking through a PIR sensor detecting region and the corresponding output signal is shown in Figure 6.

WPIR has proven it can monitor multiple targets with relative good resolution. This is promising for the SMART house application, but it needs to be able to differentiate between adults and children. It is possible for WPIR to differentiate humans and robots. This is because the signals the sensor receives are different when a robot passes the sensor and when a human does. Based on this, WPIR is able to differentiate. But since the signal of adults and parents probably will not differ much because the thermal properties are the same, WPIR needs some adjustments to be able to differentiate children from parents.

What is very promising of WPIR is that the implementation is very easy and the costs are also low. The only hardware needed is a ceiling camera which can observe the environment. Since it is also possible to monitor multiple targets WPIR can still be suitable for this application, but then the problem of differentiation should be solved. [39]

Ultrasound Localization

The ultrasound system is not always on the point of view of the today technologies, but there is no reason for this. Even nature shows that the ultrasound detection system works fine in any types of conditions. How exactly does it work? Bats have this power to orientate themselves only by using ultrasound waves that at the contact with objects return to the source. Based on the time of such a routine, the bat can detect exactly how far the target is.

This feature sounds good and it seems to work, but there is a big counter argument for using it. Especially in the case that is being addressed in this project, where the subjects are young children. This method is an invasive one, which in the long run can cause a lot of problems, especially because of its chemical effects. As stated in “The chemical effects of ultrasound”, it can drive metal particles together at such high speeds that they melt at the point of collision, and ultrasound can generate microscopic flames in cold liquids. As it is already known, in blood there are different metals like potassium, iron, calcium, mercury, sodium and many others which can react to ultrasound waves. Effects that can occur are related to headache, dizziness, and nausea, but most important one is the hearing damage. So far, the most difficult part of using such a technology is the long time exposure of subjects to the system and the age range of the subject is below 9 years. Young people are more sensitive to any types of factors compared to adults, hence it is difficult to integrate this technology into the current project analysis. The same document stated that the technology is used for industrial applications because it can radiate through large volumes of liquid, it will represent a barrier instead of a helping tool for the SMART home system. Since due to this downside ultrasound is not an option, this detection method is not further analyzed. [40]

Pressure Sensor

A pressure sensor is a device that can notice when there is put pressure on it. It can be used for this application when the sensor is placed before the staircase. The sensor can be set so, that it only sends a signal when a minimum force is put on the sensor. This way, the sensor is still able to differentiate between children and adults.

The downside of this type of detection is that a sensor cannot sense multiple people, it only senses when there is put pressure on. This way dangerous situations can occur, when for example a child walks after it parent which is going downstairs, a dangerous situation takes place. The system does not oversee the whole environment and this can cause problems. Also, the measurements can be easily tricked since the sensor only measures weight, a child can easily trick this by for example hold something heavy. These downsides lower the accuracy of the detection method and this does not have a good contribution to the system when maintaining safety is the main goal.

Active Badge

The active badge is a badge that a human wears so that he/she can be tracked by a sensing system. The Active Badge operates as a beacon, regularly signaling a unique code to a network of sensors distributed around the area to be monitored. Sightings are gathered by using a master processor which polls the sensors through a network provided for the purpose. The name and location of a badge wearer can be ascertained by looking up the badge ID in a table and looking up the location where the sighting was made. It has been continually improved on to make it more accurate. Because of these improvements, this technology is applicable for the SMART home, because it is able to track multiple people, even through walls and it is not expensive to implement. Since every user has its own ID, differentiating between adults and children should not be a problem.

A disadvantage is that the users will have to wear these badges which might be considered inconvenient by adults. On the other hand, a child might not wear it, for example when the parent forgets to put it on, or even destroy it while wearing it. Also if the badge is attached to a piece of clothing is the child able to remove that clothing. If a child is not wearing the badge, the system will not be able to track the child and the same as with the pressure sensor, the system does not oversee the full environment. These downsides bring the accuracy down to this detection system since there are a lot of bugs which can let the detection fail. [41][42]

Camera Detection

Another detecting system is camera vision. This alternative represents the best one for indoor use for detecting people. It is not invasive, which means it can also be used to detect children, it is rather cheap and efficient. But what is the state-of-the-art of such a technology? The latest cameras are connected to the internet of things which make them SMART. They are able to process via an AI all the given images and provide helpful data. The down side of it is represented by the privacy issues, but looking at the SMART house application which is the underlying thought of this project, this would not be a problem since such a house consists of multiple sensors.

Camera systems like ‘Kinect’ are able to track and observe people. This is done by software which analyzes the camera image of the Kinect camera system. This is done as showed in Figure 7.

Figure 7: Detection and tracking with camera vision

A person detector which makes use of such Kinect camera data or RGB-D is HOG (Histograms of Oriented Gradients). HOG is nowadays one of the most performant and widely used methods for visual people detection. The method considers a fixed-size detection window which is densely subdivided into a uniform grid of cells. For each cell, the gradient orientations over the pixels are computed and collected in a 1D histogram. The intuition is that local appearance and shape can be characterized by a distribution of local gradients without the precise knowledge of their position in the cell. [43]

With the HOG and the RGB-D, an image can be created where a person is detected and marked with a colored box. The height of this box can be determined and in this way adults and children can be differentiated from each other. Because the height difference between an adult and a child is big enough, the accuracy of the height of the box is not that important since the box of the adult will be way bigger than the one of the child. Therefore, the accuracy whether a person is detected or not is more important than the accuracy of the box itself. The performance of the people detection system is evaluated in terms of detection rates (accuracy) and false positives/negatives. True positives are the people images detected from the ground truth. False negatives are the people images not detected from the ground truth and false positives are images detected as people that do not appear in the ground truth. It is logical that the system will fail when the adult is not detected by the system but wants to go down the stairs, a false negative. Also, dangerous situations can appear when a false positive is detected and the child has a free entrance to the stairs. The accuracy of the detection system is then defined as follows:

[math]\displaystyle{ Accuracy=\frac{TP+TN}{TP+FP+TN+FN} }[/math]

With TP, TN, FP, and FN are respectively True Positive, True Negative, False Positive and False Negative.

It turned out that these typical camera systems have an accuracy of about 85 to 90 percent. For the system, this means that on average one in ten times the system fails and the parent has to wait or open the gate by hand. According to an article, the accuracy of this system rises if the camera uses thermal images. False positives are most commonly generated by obstacles and other objects which are placed in the room. When thermal images are used, these objects are not observed anymore since they do not generate any heat. The accuracy can rise to 92 percent or higher, depending on the resolution of the camera. This method will be elaborated later. [44]

Another advantage of using thermal images instead of normal camera view is that a normal camera fails when operating in the dark. Persons cannot be detected when the camera has lost all his vision. To cope with this problem again thermal images comes in handy, a thermal camera will not loose its functions when the environment is dark since heat is always generated by the targets.

The second problem of normal cameras is that the precise location of the target is not known. The detection system uses the 2D image to detect the targets, this can cause some problems. For example, imagine a child in front of the image and an adult in the back. When the detection system does his work, the two boxes may be of the same height because the parent is much smaller in a 2D image when it is positioned further back. The system, therefore, lost its property to differentiate properly. To still know the exact location and height of a target and prevent false positives from happening, a device which measures distances is necessary for this system. This makes the system more expensive, but the accuracy will go up together with the reliability. When the distance is known and the height of the box, the software of the system can do the rest by calculating the exact height of the person and it is able to differentiate. Also, the system can keep track of the target and see when it is near the stairs so it can act when necessary. Later more will be elaborated on this distance sensor.

Thermal Imaging

Thermal cameras are a really good way to detect people. They are often used by firefighters to detect the source of the fire or see the heat signatures of the casualties. This detection system is already in use and it was proved to be efficient, which may make it worth implementing in a SMART house. The main disadvantage is constituted by the fact that the images appear like 2D plane pictures, where the actual distances to objects are not known as stated above. For a real case situation where people can use these images to spot other persons thermal cameras perform well, but how do they cope with human detection when heat objects are around? There might occur several problems with shape detection when applied to a SMART home system for children safety. The thermal imaging shows the details of a person usually in colors of red, yellow and orange. These colors are also shown for hot objects, for example, a cup of coffee.

Thus, the main problem for the AI which shall detect persons is how exactly to distinguish between a cup of hot liquid which is close to the camera and a person which is far behind? Since the shapes are not perfectly sharp, there might be used a human body template to look for persons in such images, but another question arises: how do you identify between two persons that are on the same line facing the camera? The image will show a bunch of heat and the AI may identify one person, but it is impossible to find the second person just by using that camera. So, blind spots represent a problem for a system with only one such camera. Besides that, advantages of this detecting technology overcome most of the other ones. It is capable of detecting people with high accuracy(up to 100% if the environment details are loaded into the system and a good detection algorithm is implemented) and can help to detect them through smoke and darkness. The capability of detecting through smoke might not be needed for the purpose of the project, but the second attribute is really powerful since it can help during nights. The thermal images are also easier to process compared with regular images because they contain contrast between the heat source and regular space(usually yellow-red to green-blue). Thus, the differentiation between people and other regular objects is quite easy and efficient.

Comparing the systems

Of all the localization systems discussed above, one should be chosen to implement the system. The systems have all their pros and cons, the best system should be chosen. To choose the best system, the requirements should be addressed to make sure the system which can fulfill the most requirements is chosen. In Table 9 are all the systems adressed, with their downsides and advantages.

Table 9: Comparison of the recognition and detection systems
System Multiple targets Able to Differentiate Easy to implement Accuracy Costs Downsides Advantages
Infrared sensor Good Yes Yes Good around 97% Low The target has to be stationary and it is necessary for the sensor to be able to rotate. Very cheap with good accuracy, multiple targets is not an issue.
WPIR Good Not yet Yes Good around 95% Low Not able to differentiate yet. Low in cost, high accuracy and easy to implement.
Ultrasound - - - - - It is an intrusive method and can cause harm in long-term use, so it is not further investigated. -
Bracelets Good Yes Yes Good Medium Need to wear a bracelet all the time, hard to implement for guests. Exact location detection, the best solution to detect every person.
Pressure plate Poor Yes Yes Low Low Might be easily tricked and does not oversee the environment. Cheap alternative and easy to use.
Cameras Good Yes Yes 80 - 90% medium-high Might have dead angles, not able to work when it is dark and distance is necessary. Good vision range, easy to recognize people by using AI and easy to implement.
Thermal Imaging Good Yes Yes good around 95% medium-high Might have dead angles and it needs a second distance measurement device. Good vision range, able to function in the dark, easy to implement.

Conclusion

As described previously, the system will enhance safety for children and convenience for the parents. A system that is unreliable in 5% of the runs is not going to be agreed on by the parents. A system that fails approximately once in 20 runs is not safe for the child and is therefore not worth being implemented, especially when the production costs are pretty high.

In addition, the system must be able to track multiple people and differentiate among children and adults. Among all the possible detection solution described above, it can be concluded that the best options are IR sensors and thermal imaging. The IR sensors have a few drawbacks that the detection of the subjects might be impossible to be done, which finally leads to frustration for the parents and safety risks for their children. This method of detecting shall be able to also detect visitors and treat them like a member of the family: getting access whenever is needed and accepted by the rules of the system. The limited visibility of an IR sensor requires either multiple sensors to be used or a gimbal installed for getting an overall image of the security of the subjects. These small inefficiencies leave space for the second technology, the thermal imaging. The key points of this technology are the easy detection and differentiation between different people (all it’s drawback could be solved by installing multiple types of such cameras or few extra sensors that can provide the needed data for tracking). Compared with cameras (another possibilty to the given extent), the thermal imaging is not invading the privacy, the code is cheaper to produce and a bit more efficient compared with normal vision cameras. The accuracy is around 95% which is acceptable, but this number can grow when distances of the target can be measured so the system exactly knows where in the hallway the target is located.

Given the context of the SMART house application that shall provide automatic protection for subjects, thermal cameras are going to be elaborated further on. The design of the scenarios and the needed hardware will be adapted to the chosen detection system and the software implementation shows a possible way of detecting multiple people in the same image.

Design

Open/Closed System

There are two types of systems that can be used to solve the problem.

  • The first one is that the gate closes after the system detects that the child is close to the stairs. With “close” it’s meant that the child does not have enough time to reach the gate before it already has opened. This time would be determined by taking into account the closing speed of the gate and the average speed of a running child. After this is determined a distance can be taken that would give the system enough time which would create a danger-zone. When the child enters this zone the gate closes. However, this type of system has many issues. First of all children of different ages have different average speeds which mean that the size of the danger-zone would have to be changed constantly. The danger-zone could become so large that the system no longer is able to do its job properly. If for example the whole hallway is considered dangerous the system can no longer work properly. Secondly, the child might enter the danger zone but that does not mean it wants to go to the stairs. The child simply might walk past the stair so that it can go to a different room or the child might turn around halfway through. This means the system would constantly be closing the gate for no good reason and that should be avoided.
  • The second type is that the gate only opens after the system detects an adult. This type of system is more advantageous than the former one because it deals with most of the presented issues. The average speed of the child is no longer important and the size of the danger zone can be reduced significantly because the urgency factor is taken away. The system will also be nervous because the systems danger-zone is much smaller and it no longer has to take the erratic behavior of the child itself.

Because the second system is more advantageous than the first one it has been decided to use such a system instead.

Walking speed

The average walking speed of humans is 1.4 m/s but can go up to 2.5 m/s. The average running speed of normal humans differs from 4.4 m/s to 6.7 m/s. Since the subject in this project is a child, these numbers are way too high. For the children group from 1 to 4 years, the approximated average maximum speed is about similar to the normal walking speed of a human, so about 1.5 m/s (which in most cases will not be reached). For the age group 5 to 9 years this speed gradually increases the older they get, but will not rise above the average running speed of normal human. So this maximum value can be approximated at 2.5 m/s at the age of 9. Due to certain factors such as education during the early years and the small distances the hallways provide, these speeds are most of the time not reached around the staircase. Especially considering the time for accelerating and decelerating the maximum distance they can travel in a time window is a lot lower than their actual maximum. An estimation can be done, which has a safety margin, at 2 m/s for the older children. The gate should be quick enough to fully close in time when the child enters the ‘danger zone’, and the time the gate takes for closing should be less than the time it takes for the child to cross the danger zone. [45] [46]

Speed of gate

The speed of the gate is dependent on the motor that is going to be used and the amount of RPM (Rotations per minute) the motor can reproduce. With the use of a matlab script some calculations have been done. With some average motors that can rotate at an rpm of 150 to 200 the opening and closing speed of the gate can be approximated to be somewhere between 0.5 and 1.5 seconds. For this design, a motor will be chosen that has at least 180 rpm, so that the opening/closing time is a little bit more than 1 second. For the gate material something light weighted is prefered. For the calculations a rectangular tube profile is used for the beams of the gate, with the dimensions 25x25x750. In the Matlab script, the torque that is needed is approximated at 1 Nm. Thus the motor should have a torque of at least 1 Nm. This is a motor that would fit the bill.

Danger Zone

The maximum speed of a child can be approximated to be around the 2 m/s. But since the child will not be running on its maximum speed the whole time, the estimated average speed of the child is therefore set at 1.2 m/s. The dangerzone has to make sure that the child can never come between the gate and the opening, therefore also a safety margin has to be added. Together with the opening and closing speed of the gate, the dangerzone radius is chosen at 1.5 m. The LED lighting to make situations visible, warn the parents and get the attention of the parents will be placed in the top of the gate, pointing upwards. This means these lights are always visible, even when the parent is carrying something that blocks his/her view to the floor.

Recognition and detection

The detection and recognition software shall be able to process given thermal images. As a result of processing, the human bodies will be discovered from the given image and a possible differentiation between a child and an adult follows. The image is given in a RGB format, every pixel taking three value between 0(minimum) and 255(maximum) for the red, green and blue details of the pixel. These colors have the purpose to represent heat details of the image: from blue(cold heat signature) - green - yellow - orange - red(hot heat signature). The colors are correlated to the reality heat temperatures, so by using the contrast between colors it becomes easy to detect a person when no heat source is close. The processing is done for every pixel one by one where the color is detected and in case it is between the threshold values set by us, it will be converted to a white pixel in a new created image. Finally, the image created contain only white areas that represent the heat signatures from the given thermal image. Now, the detection part has to come in handy and be able to distinguish which part of the image belongs to which person. The detection is done by saving the maximum and minimum X and Y values of the shape. If the distance between the pixels is larger than a threshold given by us, the program shall consider a new person is present in the picture. Of course, it is consider as being a valid person if these values make sense and the height measurement > width measurement of the shape. In order to process images, OpenCv library is used for C/C++ code. Even if the image processing looks easy and fast, it takes a few seconds before the task completes because every pixel shall be analysed. For a full hd image it is needed to compute 1920x1080 pixels values. Examples of what the code does, can be seen in Figure 8 and Figure 9.

Figure 8: Thermal imaging software
Figure 9: Thermal imaging software

Source code: File:Source.zip


The working of the system

The system will use a “gate closed” principle. Thus the gate will always be closed unless the parent is detected in the ‘danger zone’. The system is detecting the position of the subject in the hallway constantly (when movement is detected). So when the parent is moving towards the stairs it will already calculate time and its actions. When both a parent and the child are detected in the danger zone, the LEDs change to an alarming and to a striking preference color. This also means that the gate will not open when the child is in this zone, due to safety precautions. So when the parent is walking towards the gate and comes in a certain proximity of it, and there is no child in the danger zone radius, then the gate will open automatically and the parent can just walk through the gate and down the stairs without any inconvenience, and when the parent is on the stairs, the gate will close behind them.

Gate design

At the top of your stairs only use hardware-mounted gates. Pressure-mounted gates are ideal for between rooms or at the bottom of the stairs, but they are not strong or safe enough for the top of the stairs. If you choose an accordion-style gate, make sure it has a top filler bar. If it does not, the child may get his fingers, hands or head caught in the spaces. The gate should only move in one plane and should be installable for every stair casing. The design of the gate then will be a casing with the height of the gate, that can be mounted on the wall or on the stairway railing itself. On the opposite site of the mounted casing, there will be a rail where the gate shall fit in, and this will make the construction stronger and so that it cannot be pushed out. This rail and the end of the gate are finished with a soft, child-friendly material for the safety of the child when being near the (moving) gate.

The gate design requirements are determined and a first design sketch was made. In this overview of the hallway and its setup the gate is described pretty concisely and is not yet specified. To show the way the gate is going to work, look and be placed in the setting, a CAD design of the gate is made. In this design, the materials are not specified and how it is going to look eventually can be different as well. The mechanism and the operating are visualized and the different parts are assembled. Below is the design and the mounting in a stairway visualized. The front view of the gate can be seen in Figure 10 and the back view of the gate can be seen in Figure 11.

Figure 10: Front view of the gate
Figure 11: Back view of the gate
Figure 12: Exploded view gate design
Figure 13: Transparent gate design showing the mechanism inside

The exploded view of the different components, these will get a further explanation in a components list in the future. On the right, you can see a frame view of the mechanism when it is folded in so that the gate is open. The plate on the front of the gate is representing a sheet of textile that is strong but soft to protect and prevent anyone from touching the mechanism from the hallway side. The exploded view of the gate can be seen in Figure 12, the transpart gate design showing the mechanism inside can be seen in Figure 13.

For the setup of the gate when it is applied in a staircase it is shown mounted on a wall on an average staircase and hallway for example. See Figure 14 and Figure 15 for gate closed situations and Figure 16 and Figure 17 for gate open situations.

Figure 14: Front view of the gate closed in an example hallway
Figure 15: Back view of the gate closed in an example hallway
Figure 16: Front view of the gate opened in an example hallway
Figure 17: Front view of the gate opened in an example hallway

Camera specifications

The camera needs to satisfy certain needs:

  • Has to be mountable on the ceiling (2.5 meters high)
  • Needs to have a clear view of the entire hallway
  • Needs an FOV of at least 90°
  • Minimum range of 6 meters
  • Quality of images needs to be good

If a camera satisfies these needs than this camera could be used for a SMART house. A couple camera’s that already exist and satisfy these needs are:

A 360 camera needs to be centred in both the hallway and the staircase. The hallway has a width of 0.95 meters so the 360 camera should be 0.45 meters away from both the walls. Because the width of the staircase is variable for different houses the centre would have to be measured by the user or the professional that installs the system. The 90 camera must be positioned so that the whole danger zone is viewable. The sketch shows the positions for each hallway where such a camera can be placed so that there are minimal dead angles. With a reach of 6 meters this camera is able to see most, if not all in some cases, of the hallway. By positioning this camera in a smart way no parts are needed that make the camera rotate.

Sensors

Figure 18: A sketch of the situation describe to measure the height using the second type of sensor

For an IR camera to be able to determine the difference between a child and an adult the system will need to know the distance of the person to the camera. An IR camera is unable to determine distances so additional sensors are needed. There are several sensors that are able to use IR light to determine the distance from that sensor to the object it is measuring. The first type of sensor uses a class 1 IR laser to measure distances. This laser can only be directed in one direction which means multiple would have to be used in order to be useful for the system. There is also the added problem with these kinds of sensors that they cannot be positioned easily in a hallway so that the measurement can be used by the system to eventually determine the height. Therefore this type of sensor would not work. The second type of sensor that could be used is very similar but it has a 360 degrees viewing angle which allows for flexibility in placement. These type of sensors can be used to measure the height of a person that is walking through them if you mount them on an angle a. When a person walks through this sensor it will start to measure the distance L, the distance from the sensor to the person. This distance L will decrease if the person walks deeper into the danger zone and it will decrease if the person walks out of the danger zone. When the person has fully entered the danger zone the last measurement data of the sensor can be used to determine the height of this person using the following equation:

[math]\displaystyle{ h = H-y = H-L\cos{\pi/2 - a} }[/math]

where H is the height of the hallway. Figure 18 shows a sketch of the situation. If this height h is too low that would mean a child is entering the danger zone and if this height is high enough it means an adult is entering the danger zone. The system is, therefore, able to determine the difference between child and adult. The system, however, does need to keep track of who’s inside the danger zone. This means that the system has to remember who entered it and who exited.[47] [48]

Both sensor type 1 and 2 use a laser of class 1 and although these lasers are considered eye safe it could be of a concern for certain people.

An alternative to these two types would be a sensor that uses LEDs instead of a laser, which would take away any viable concerns. These type of sensors are mounted on the ceiling and directed straight down and with a maximum measuring distance of 2 meters for black surfaces, they would still be sufficient for use in a hallway. It would have to be more accurately determined if this 2 meters still accurate if used on a black shirt or black hair to determine distances. For a white surface, the maximum distance is 8 meters. They have a measurement angle of 88 degrees which, for a hallway of 2.5 meters, would mean that you have a blind spot in the 2 upper corners of the hallway. This is not a problem however because these blind spots would have a height of about 1 meter which means that it can still measure if a person is shorter than 1.5 meters, even if this person skirts alongside the wall.[49]

All 3 types of sensors use Pulse Ranging Technology to determine distances. With this measurement method, a powerful light source emits short, high-energy pulses, which are reflected by the target object and then recaptured by a light-sensitive receiver. During this process, the emission and reception times are detected with a high degree of precision. From the values determined, the distance to the target object is calculated using the runtime of the light pulses. If the target object is close, the light propagation time is short. If the object is further away, the light propagation time is longer. The main disadvantage these sensors suffer from is that they use height to determine the difference between a child and an adult. If a child raises himself such that could make the system believe it is taller than it actually is, he might lead to the opening of the stairgate.[50]

Figure 19: (a) A schematic presentation of a PIR sensor with dual sensing elements aligned in a motion plane. (b) It's output signal when walking.

Another way to determine the difference between a child and parent would be by using a thermal camera and PIR sensors. First, the size of the danger zone would be determined and that size can then no longer be changed. A PIR sensor is positioned on the ceiling on the every edge of the danger zone. This means you would need about 3-4 PIR sensors. The PIR sensors will tell the system when a person enters or leaves the danger zone. Because the distance between the thermal camera and the edges are set, using a height calibration, a person that walks through one of the edges can be measured. And if this person is tall enough the stairgate will open. An added benefit to this system is that the thermal camera can continue to track people inside (and outside) of the danger zone which makes it easier to keep an eye out on children. PIR sensors are not expensive which is also a benefit. PIR sensors also allow to divide the edge up in zones. This means that a person can be more accurately detected when they pass through the sensor. This can also help the camera, because the camera can only see in 2D. This can lead to problems if two people are in the hallway but one of them walks through the sensor while the order stands further in the background. The camera will think these 2 people are standing next to each other. But because the PIR sensor can determine which of these 2 people walked through a part of the sensor that information can be used by the camera to determine who walks into the danger zone and who is not. In order for this to work, a thermal camera is needed and current technology already has 360 thermal cameras. A camera that is mounted on a rotating surface would work as well and there are plenty available.[51] [52] [53]

In conclusion, the first type of sensor does not work for the SMART home system. The second type could work if the system is able to remember who is inside the danger zone and if the class 1 laser that is used is eye safe, under any condition. The third type of sensor, using IR LED’s, has two disadvantages. Namely that children can pretend to be taller than they actually are and that it has only a maximum range of 2 meters for black surfaces. This means that a child could crawl under this sensor and not be noticed by the system. This could be solved by installing a PIR sensor in the danger zone. This sensor can then check if the danger zone is indeed empty. For this type of sensor, the system also needs to remember who enters the danger zone just like sensor type 2. The best detection type would be by using a thermal camera and multiple PIR sensors. The main disadvantage of this system is that it is more expensive than the others because thermal cameras are expensive. One issue that all these systems have is the situation of when somebody comes up the stairs and somebody is also in the danger zone. The system might make the mistake that the person that is coming upstairs is the person inside the danger zone and then would close the stairgate even though it should stay open. One way to solve this is by installing a PIR sensor onto the stairs because PIR sensors can detect the direction of motion of a person(See Figure 19). [54]

Software code details

The image processing starts from top left corner and the pixels are transformed to white pixels in a new image if they are within the boundaries. The new pixel image can be clear or noisy, so different adjustments like eroding/dilating pixels are needed. The minimum and maximum X, Y are saved as described above and their difference (Xmax - Xmin, Ymax - Ymin) is tested to be greater than a normal person dimensions such that no external heat can affect the detection. In the end, a rectangle is drawn by using these 4 coordinates. The combination of this camera and a distance measurement sensor it is possible to determine the position of the persons in the room and their size, such that the differentiation can be done between a child and an adult. In the end, the algorithm shows two images: the original image where a rectangle is used to surround each subject and a black-white image that shows the heat signatures that are considered in building the rectangles. If the temperatures decrease (see Figure 20), the algorithm cannot distinguish between the human and environment anymore, so only the heat parts are considered for detections.

Figure 20: Multiple persons detection

Scenario's

Table 10: Scenario 1
Specifications
Type of hallway L-shaped
Who A child and an adult
Where Both in the hallway. The child is not in the dangerzone.
What is the plan of action The child is playing somewhere in the hallway out of the dangerzone, the adult enters the hallway and wants to go downstairs.
What is the plan of the system The infrared camera has detected the child. The sensors have not yet detected the child, so the system knows that it is not in the dangerzone. The infrared camera then picks up a new heat source. The sensors also detect that someone has walked passed them. The system then distinguishes that it is an adult that has entered the dangerzone. Since the child is not in the dangerzone, the system’s action will be that the gate will be opened.

The scenario plays in an L-shaped hallway. The 90° camera is mounted in the bottom left corner so it is able to see everything in the hallway that is of importance to it. The sensors are mounted in the ceiling at the positions that define the dangerzone, in this case there are 3 sensors necessary. For the dangerzone, see the red dotted lines in the figure.

There is a child in the hallway. The infrared camera has detected a source of heat, but the sensors have not detected someone passing them. The sensors specify the dangerzone, therefore the child is not in the dangerzone and not of importance.

There is an adult entering the hallway and moving towards the stairs. The infrared camera detects a new source of heat. After a couple of seconds, the sensors also detect that someone passes them. At this point the infrared camera compares the thermal image with its calibration. The system then recognizes the one in the dangerzone to be an adult. Since only an adult is inside of the dangerzone, the action of the system is to open the gate and let the adult pass. When the adult has passed the gate, the system automatically closes the gate again.

Figure 21: Scenario 1
Table 11: Scenario 2
Specifications
Type of hallway Cross-shaped
Who A child and an adult.
Where Both in the hallway. The child is in the dangerzone.
What is the plan of action The child is playing in the hallway, the adult enters the hallway and wants to go downstairs.
What is the plan of the system The system has already detected and distinguished the child, so it is already tracking him/her. When the adult enters, the system detects a new source of motion and has to detect and distinguish this to be an adult. Then it has to track the movement of both the child and the adult. Since the child is in the dangerzone, the system cannot open the gate when the adult enters the dangerzone. Therefore the system has to warn the adult using the LED signals that the child is also in the dangerzone and it cannot open the gate. The adult can then either choose to open the gate manually or to have the child move out of the dangerzone.

The scenario plays in a cross-shaped hallway. The 90° camera is mounted in the bottom right corner so it is able to see everything in the room that is of importance to it. The sensors are mounted in the ceiling at the positions that define the dangerzone, in this case there are 3 sensors necessary. For the dangerzone, see the red dotted lines in the figure.

There is a child in the hallway. The infrared camera has detected a source of heat. The sensors have detected someone passing them. The sensors specify the dangerzone, therefore the child is in the dangerzone and is of importance to the system.

There is an adult entering the hallway and moving towards the stairs. The infrared camera detects a new source of heat. After a couple of seconds, the sensors also detect that someone passes them. At this point the infrared camera compares the thermal image with its calibration. The system then recognizes the new person in the dangerzone to be an adult. Since only both the child and the adult are inside the dangerzone, the action of the system is to alert the adult that there is a child in the dangerzone with the help of the LED’s. Since the system cannot ensure safety for the child when it is in the dangerzone. Therefor the adult has two options, either make sure that the child moves out of the dangerzone or open the gate manually and be responsible for the consequences of that. When the adult has passed the gate, the system automatically closes the gate again.

Figure 22: Scenario 2


Table 12: Scenario 3
Specifications
Type of hallway All
Who An adult.
Where In the hallway.
What is the plan of action The adult has entered the hallway and is moving towards the gate.
What is the plan of the system The thermal camera detects a source of heat. When the sensors also detect that someone passes them, the system will distinguish the heat source that passed the sensors and sees that it is an adult. When the system has recognized the person to be an adult and knows that it is inside of the dangerzone. The system then knows that there is no danger, so the gate can be opened.


Table 13: Scenario 4
Specifications
Type of hallway All
Who A child.
Where In the hallway.
What is the plan of action The child has entered the hallway and is walking around and playing.
What is the plan of the system The thermal camera detects a source of heat. When the sensors also detect that someone passes them, the system will distinguish the heat source that passed the sensors and sees that it is a child. When the system has recognized the person to be a child and knows that it is inside of the dangerzone. The system then knows that there is a possible danger, so the gate will stay closed.


Table 14: Scenario 5
Specifications
Type of hallway L- and U-shaped.
Who An adult.
Where Coming out of the room the closest to the stairs.
What is the plan of action The adult is coming out of the room and wants to go downstairs.
What is the plan of the system The door the closest to the stairs are in these kinds of hallways already inside of the dangerzone. The system detects motion and has to detect and distinguish this to be an adult. Since the motion is already detected inside the dangerzone, the system will not be able to have the gate open in time. Therefore the gate starts opening at the moment that the system has distinguished this to be an adult. In this scenario the adult may have to wait a little bit for the gate to be fully opened.


Table 15: Scenario 6
Specifications
Type of hallway All.
Who An adult and child.
Where Child is entering the dangerzone while the adult is already in the danger zone.
What is the plan of action The adult is already in the danger zone and the gate is already opening. The child then enters the dangerzone.
What is the plan of the system The system has already detected and distinguished the person inside of the dangerzone to be an adult and knows that there is someone else in the hallway, but not in the dangerzone. Therefore the system opens the gate. While opening the child enters the dangerzone and the system distinguishes it as a child as well. At the moment the system sees the child in the dangerzone and knows that the gate is opening, the LED’s will go on to warn the parents. Further actions of the system depend on how far the gate has opened. When the gate is at least half opened, the system will notify the parent with the LED’s but will keep opening the gate. In this case, the safety of the child is in the hands of the parents. When the gate is not opened at least half, the system will notify the parent as well, but will close the gate. Therefore the parent is able to either command the child to move out of the dangerzone, carry the child out themselves or use to manual opening. In case of manual opening, the safety of the child becomes the parent's responsibility again.

Further Improvements

Hardware

The proposed system has multiple shortcomings which need improvements:

  • The first shortcoming is that the system cannot measure distances. It only knows the size of the danger zone itself and is then able to measure people on the borders of the zone but outside of the zone that is not possible. If distances can be accurately measured the system can detect people everywhere in the hallway which would make the system a lot more reliable. The system could then also be more easily implemented in other parts of the house where it’s harder to use a fixed size for the danger zone. There are sensors that can measure distances in a human-friendly way but the range of these type of sensors is lacking. There is no real incentive yet to develop such sensors because most of these sensors are used in the industrial field so human-friendliness is not a big priority. It is advisable that more research in this field is done so that human-friendly sensors can be developed.
  • Another problem is that the danger zone is some hallways is smaller than desired because of the hallway’s shape. For example a door could be inside of the danger zone which would mean that the distance between this door and the stairs is too short and that would put the child at risk. These rooms are essentially huge blind spots for the detection system. If the detection system is implemented in every room of a SMART house this problem would dealt with.
  • Currently a casing of 20 cm is used to house the motor and the folded gate. If the size could be reduced more space would be left over for the users. This is desirable because hallways already are tight so every centimeter counts. The size of the casing could be reduced if the motor that is used would be smaller. It is expected that these type of motors will continue to shrink and still will be able to deliver the same amount of torque.
  • High quality thermal cameras are also expensive which is a big disadvantage. However over the years thermal cameras have become more accurate and cheaper which bodes well for the future of thermal imaging. The expectation is that these shortcomings will be dealt with in the near-future because all these shortcomings are technical related and history showed that those type of shortcomings can be overcome. [55]
  • It is also needed that actual tests using this system are done to determine the effectiveness of this system. These tests could find more potential issues that would have to be dealt with.

Software

Figure 23: Thermal detection of a cup and a person

The localization is done as described above, but some downsides are present as it is expected for new technologies. If the subject stays close to another heat object, their shapes cannot be separated in code anymore since no clear way to distinguish which pixel each of them owns. This part of the algorithm shall be improved since the camera can detect you and your kid like one subject and behave abnormally without any reason. The solution to this problem can be solved with the distance sensor that can say which pixels are from which person and who is in front of whom.

A second improvement shall be developed in the person detection. In the example below, a cup of coffee is left on a table close to the thermal camera and the subject is far behind in the environment.

Since the current person detection is done by finding few black lines of vertical pixels between objects, any heat source that fits in the dimensions requirements can be misinterpreted as being a human. This ensures that the subject is connected with all his/her parts(even if they are few pixels apart from the main body, hands, legs and other small parts of the human body are considered as belonging to that person. The problem could be used by applying a modified BFS which shall test about 5 pixels in all directions, but it would be heavily inefficient since the complexity for a BFS with a huge number of possible neighbors increases exponentially.

Expansion Towards Other Rooms

The underlying idea of this project is the SMART house. The goal is to take the idea of a SMART house and research how to improve the child safety of it. It was found that the most child related accidents happen inside the house due to falling. Therefore the possibility to implement the SMART house idea to have a child safe stair is looked at. The next step is to see how this child safe stair can be implemented in the rest of the SMART house idea. Since a house contains more rooms that consist of potential danger for a child, for example the kitchen and the bathroom, there is looked at how the child safe stair system can be converted to other rooms.

The system is able to differentiate children from adults, this is the crucial part of our system. If the system is implemented in other rooms as well, this software can still be used to differentiate. There is however one remark to make. For the stairs, research have been done on what the best system hardware is for this particular problem. For this problem an IR camera was apparently the best option, the differentiation software is therefore created with the output of the IR camera. It could be possible that for other rooms a different type of recognition system should be used. For example, it could be possible that for the bathroom IR is not needed and therefore unnecessary to implement for this application. In this case the differentiation also have to be done in a different way.

Looking to the investigated child safe stair system, implementation opportunities are mostly limited to the kitchen. Since in the stair problem the only interesting thing for the system to sense is the child and its position in the hallway, IR is a great method to use because it filters out all the objects around the target. But this does not apply to the kitchen or the bathroom, because dangerous objects like electrical sockets, knifes and sinks must be detected by the system. Also the problem the system solves, which is reducing injury due to falls, is not applicable here. It would be impossible for such a system to detect if for example a child climbs on a chair and recognizing that this is a dangerous situation. However, the kitchen does have potential dangerous devices which radiate heat, like the oven or the stove. For these kinds of problems the investigated system can be converted so that it also works in these other environments. This way the system can differentiate adults and children with the used software. Some adaptations should be included because with this software, all the object which radiate heat are seen as a person. But this could be easily solved by implementing the property that a person is able to move. When the differentiation is successfully performed, the next step is to recognize potential dangerous situations. In this case, a dangerous situation is when the child comes too close to an object that is too hot. When the system does detect this, an action has to be performed. This could be for example shutting down the stove, but this will not take the heat away, or warn the parent so they are aware of the situation. In this way it would be possible for the system to detect when the child is in a potentially dangerous situation. For example when there is a cup which contains a hot liquid on the table. In this way the child safe stair system can be implemented in the SMART house. More research has to be done to find out whether a room needs an IR sensing system. When this is the case, the software with some adjustments can be used to differentiate adults from children. The system would be applicable for the kitchen already, but also here more research has to be done to come with a final working system.

USE Aspects Child Safe Staircase

In this part, there will be recapitulated to the USE aspects defined in the beginning of the report about the whole implementation of child safety in a SMART house. In the beginning, the USE aspects have been defined for all the actors and how they will benefit from the first idea. This idea was to implement child safety in a SMART house with use of intelligent systems. During the project this idea is narrowed down to just the staircase and how to make this more child safe and implementable in a SMART house. The question rises whether this child safe stair case system, which is designed in this project, can still be beneficial for the USE actors defined in the beginning of this project and how. Because there is looked how to implement such child safe systems in the rest of a SMART house, this question is key for further development.

Users

The primary, secondary and tertiary users as defined in the beginning will not change. The system will be used in the same way as any other child safe system in a SMART house will do, so narrowing down to just the stairs did not have any affect. For the primary users, mainly for the parents, the system is the most beneficial. The system will reduce stress because the child will not be in danger while at the stairs and the convenience of using the stairs with a stair gate increases. This is also true for the secondary users.

Society

Society can also still benefit from this project because of the better health statistics. There is no research done if it lowers the parental stress, only a speculation can be made. But the child safe stair system is able is indeed able to lower the accidents which happen at the staircase among children. It is proven that state-of-the-art baby gates often does not function as properly as they are designed for. The incidence of gate-related injuries increased significantly from 3.9 per 100.000 children in 1990 to 12.5 per 100.000 children in 2010. Most often injured by falls down stairs after the collapse of the gate. The system designed in this project should prevent these accidents from happening. Also injuries by contact with the gate which results in open wounds and soft-tissue injuries. These kind of injuries should also be prevented with this system because the possibility to get stuck or jammed between the gate is made as low as possible.

The estimated impact of the system can therefore be calculated. Annually, an average of 93.189 injuries per year or 46.5 injuries per 10.000 population happen among children due to stair-related accidents. Of this number, an estimated 37.673 accidents happen because of injuries associated with gates due to the above explained accidents. The gate will not be able to collapse and open wound injuries will also be prevented as much as possible. Looking to the state-of-the-art child proof equipment, which often is functioning very poorly, this system will reduce injuries which are cause of failing of these kind of equipment. With these numbers an estimation can be made about how many injuries the system will prevent. Looking at the numbers, the system will prevent an estimated amount of 50.000 injuries annually. This number is based on the accidents which happen due to collapsing of the gate, the number of stair related accidents and also the chance of opening gates due to unreliable child proof equipment. This number is based on data of the USA, so the number of 50.000 prevented injuries per year is only for the USA. In the accident research there is seen that the number of accidents which in the USA and the EU happen approximately the same. Therefore this estimated number will approximately double when looking to the USA and EU. This is a significantly number and there can be stated that the system has a significant impact when used. This proves that society can definitely benefit because of the better health statistics the system will bring.

Enterprise

All the aspects discussed in the beginning how enterprise can benefit from a child safe SMART house system also counts for this intelligent subsystem. There will be increased products sales because there will be new products to sell and the pay-as-you-go house service can be implemented for the staircase. Also research still have to be performed because the concept is not fully developed yet. In the section ‘Further Improvements’ there are some problems stated which the system was not able to solve. To solve these problems research is necessary and the outcome of this research can be profitable for the enterprise.

Conclusion and Evaluation

The idea of implementing child safety into a SMART House has a lot of possibilities and offers a variety of choices to keep your child more safe and be convenient for the parents at the same time. Since the SMART House itself is still a project that is under development, it is still very hard to implement things into already.

This project tackled the problem of falling down the stairs and to reduce related accidents. The way to fix this was to improve on the already existing stairgates. Automating the gate improves a lot to both the safety of the child as well as the comfort of the parents. However this system only works for simple, basic situations as off yet.

Although it is possible to make an automated stair gate, and being able to detect and distinguish a child from an adult, the system still has a lot of shortcomings. These shortcomings make the system not very trustworthy. But all of these shortcomings are problem that are fixable, just not with the technology that is currently available. Another problem of the system is that, in only the staircase problem, it has to be cheap for the system to be interesting for users. Since the system needs state-of-the-art technology for it to work, the price is still an issue. But since the SMART House is still under development and it will likely take a couple of years till it can and will be used, this problem will likely be solved at the time.

References

  1. ↑ https://www.forbes.com/sites/freddiedawson/2015/11/27/how-to-make-money-from-smart-homes/#4ff6084e4235
  2. ↑ https://www.verywell.com/common-causes-of-stress-for-mothers-3144845
  3. ↑ http://raisingchildren.net.au/articles/stress_management.html
  4. ↑ https://en.wikipedia.org/wiki/Externalizing_disorders
  5. ↑ Costa, N.M., Weems, C.F., Pellerin, K. et al. J Psychopathol Behav Assess (2006) 28: 113.doi:10.1007/s10862-006-7489-3
  6. ↑ Jones, H.A., Putt, G.E., Rabinovitch, A.E. et al. J Child Fam Stud (2017) 26: 225. doi:10.1007/s10826-016-0547-x
  7. ↑ Bayer, J. K., Sanson, A. V., Hemphill, S.A. et al. J Applied Developmental Psychology (2006) 27: 542. doi:https://doi.org/10.1016/j.appdev.2006.08.002
  8. ↑ Cooklin, A. R., Giallo, R. and Rose, N. (2012), Parental fatigue and parenting practices during early childhood: an Australian community survey. Child: Care, Health, and Development, 38: 654–664. doi:10.1111/j.1365-2214.2011.01333.x
  9. ↑ https://www.rtlnieuws.nl/gezin/ouders-zijn-te-voorzichtig-laat-kinderen-meer-risicos-nemen-met-spelen
  10. ↑ https://www.nrc.nl/nieuws/2017/03/31/advies-laat-kind-met-zakmes-spelen-7799114-a1552801
  11. ↑ https://www.psychologytoday.com/blog/the-power-prime/201403/risk-taking-your-children-how-much-is-enough
  12. ↑ Oldehinkel, A. J., Veenstra, R., Ormel, J., De Winter, A. F. and Verhulst, F. C. (2006), Temperament, parenting, and depressive symptoms in a population sample of preadolescents. Journal of Child Psychology and Psychiatry, 47: 684–695. doi:10.1111/j.1469-7610.2005.01535.x
  13. ↑ Heider, D., Matschinger, H., Bernert, S. et al. Soc Psychiat Epidemiol (2008) 43: 266. doi:10.1007/s00127-007-0302-0
  14. ↑ Nishikawa, S., Sundbom, E. & Hägglöf, B. J Child Fam Stud (2010) 19: 57. doi:10.1007/s10826-009-9281-y
  15. ↑ Overbeek, G., ten Have, M., Vollebergh, W. et al. Soc Psychiat Epidemiol (2007) 42: 87. doi:10.1007/s00127-006-0115-6
  16. ↑ Martina K. Gere, Marianne A. Villabø, Svenn Torgersen, Philip C. Kendall, Overprotective parenting and child anxiety: The role of co-occurring child behavior problems, Journal of Anxiety Disorders (2012), 26: 642, https://doi.org/10.1016/j.janxdis.2012.04.003
  17. ↑ https://www.forbes.com/sites/freddiedawson/2015/11/27/how-to-make-money-from-smart-homes/#13c31bfb4235
  18. ↑ https://www.forbes.com/sites/aarontilley/2016/07/06/vivint-smart-home/#421511b5525e
  19. ↑ http://internet-of-things-innovation.com/insights/uncategorized/impact-of-smart-home-technology/#.WQnYjuWGPb0
  20. ↑ http://www.rospa.com/home-safety/advice/child-safety/accidents-to-children/
  21. ↑ http://injuryprevention.bmj.com/content/injuryprev/2/4/290.full.pdf?sid=0e46d567-3cc5-4180-9154-056b903f6cbc
  22. ↑ https://academic.oup.com/eurpub/article-lookup/doi/10.1093/eurpub/ckq047
  23. ↑ http://www.euro.who.int/__data/assets/pdf_file/0003/83757/E92049.pdf
  24. ↑ http://www.eurosafe.eu.com/uploads/inline-files/IDB_Report_2014_final%202010-2012.pdf
  25. ↑ Stair-Related Injuries to Young Children Treated in US Emergency Departments, 1999–2008 Ashley E. Zielinski, Lynne M. Rochette, Gary A. Smith Pediatrics Apr 2012, 129 (4) 721-727; DOI: 10.1542/peds.2011-2314
  26. ↑ https://en.wikipedia.org/wiki/Childproofing
  27. ↑ https://en.wikipedia.org/wiki/Child-resistant_packaging
  28. ↑ http://abcnews.go.com/blogs/health/2013/04/25/child-resistant-pill-bottles-easily-defeated/
  29. ↑ https://www.nps.org.au/australian-prescriber/articles/when-is-child-resistant-packaging-not-child-resistant
  30. ↑ http://www.pharmacytimes.com/publications/issue/2007/2007-05/2007-05-6518
  31. ↑ M.J.Russell, Determining if Child Proof Containers Are Really Child Proof: Are Child Proof Containers Really Safe from Children? (2009)
  32. ↑ Gabriel H.C. Bonfim, Fausto O. Medola, Luis C. Paschoarelli, Correlation among cap design, gripping technique and age in the opening of squeeze-and-turn packages: A biomechanical study, International Journal of Industrial Ergonomics, Volume 54, July 2016, Pages 178-183, ISSN 0169-8141, https://doi.org/10.1016/j.ergon.2016.06.004.
  33. ↑ A. Yoxall, E.M. Rodriguez-Falcon, J. Luxmoore, Carpe diem, Carpe ampulla: A numerical model as an aid to the design of child-resistant closures, Applied Ergonomics, Volume 44, Issue 1, January 2013, Pages 18-26, ISSN 0003-6870, https://doi.org/10.1016/j.apergo.2012.04.006.
  34. ↑ http://www.irishexaminer.com/ireland/1-in-4-parents-admit-child-opened-door-of-moving-car-314021.html
  35. ↑ http://allaboutchildsafetygates.blogspot.nl/2009/01/three-types-of-child-safety-gates.html
  36. ↑ https://www.ncbi.nlm.nih.gov/pubmed/24530221
  37. ↑ Mrinal Kanti B, Thermal infrared face recognition - a biometric identification technique for robust security system, Intech, 2011
  38. ↑ Barshan B. Taget differentiation with simple infrared sensors using statistical pattern recognition techniques, ScienceDirect, 22 March 2006
  39. ↑ R. C. Luo and O. Chen, "Wireless and Pyroelectric Sensory Fusion System for Indoor Human/Robot Localization and Monitoring," in IEEE/ASME Transactions on Mechatronics, vol. 18, no. 3, pp. 845-853, June 2013. doi: 10.1109/TMECH.2012.2188300
  40. ↑ http://www.scs.illinois.edu/suslick/documents/sciamer8980.pdf
  41. ↑ R. Want and A. Hopper, "Active badges and personal interactive computing objects," in IEEE Transactions on Consumer Electronics, vol. 38, no. 1, pp. 10-20, Feb 1992. doi: 10.1109/30.125076
  42. ↑ Y. Zhao, N. Patwari, P. Agrawal and M. Rabbat, "Directed by Directionality: Benefiting from the Gain Pattern of Active RFID Badges," in IEEE Transactions on Mobile Computing, vol. 11, no. 5, pp. 865-877, May 2012. doi: 10.1109/TMC.2011.89
  43. ↑ L. Spinello, "People Detection in RGB_D Data", in International Conference on Intelligent Robots and Systems, September 2011
  44. ↑ L. Suspereggi, "On the Use of a Low-Cost Thermal Sensor to Improve Kinect People Detection in a Mobile Robot ", Sensors, 2013
  45. ↑ https://www.quora.com/What-is-the-average-running-speed-of-a-human
  46. ↑ https://en.wikipedia.org/wiki/Preferred_walking_speed
  47. ↑ https://www.sick.com/de/en/distance-sensors/mid-range-distance-sensors/dx35/dt35-b15851/p/p309055
  48. ↑ http://www.pepperl-fuchs.nl/netherlands/nl/2274_entfernungsmessger_te_distanzsensoren.htm?view=productdetails&prodid=73695
  49. ↑ http://www.pepperl-fuchs.nl/netherlands/nl/classid_53.htm?view=productdetails&prodid=62235#documents
  50. ↑ https://www.pepperl-fuchs.com/global/en/23466.htm
  51. ↑ https://en.wikipedia.org/wiki/Passive_infrared_sensor
  52. ↑ http://www.thermalradar.com/thermalradar/
  53. ↑ http://www.360visiontechnology.com/visiondome
  54. ↑ J. Yun and M. H. Song, "Detecting Direction of Movement Using Pyroelectric Infrared Sensors," in IEEE Sensors Journal, vol. 14, no. 5, pp. 1482-1489, May 2014. doi: 10.1109/JSEN.2013.2296601
  55. ↑ http://apfmag.mdmpublishing.com/history-thermal-imaging-innovation/
Retrieved from "https://dsdwiki.wtb.tue.nl/index.php?title=PRE2016_4_Groep1&oldid=44839"