Tecnalia is working with Ibermatica in Computer Vision technologies for intelligent algorithms on embedded systems. These algorithms allow the detection of people falls at home or, in general, still people for long time in no regular zones. When one of these incidences is detected the algorithm must create an alarm signal to inform to the main control.
This intelligent algorithm for incidences detection is built with two different operational modes:
- Training mode: In this mode it is necessary to place the system for enough time to train the room. During this period of time the algorithm will learn by means of images the user habits in terms of what are the room places where the user spends most of his/her time. The algorithm classifies the room area in three different zones: regular zone, transition zone and irregular zone.
- Monitoring mode: In this mode the system will use the trained zones to resolve when a user is in a regular zone or not. When a user is detected entering in a monitored zone, a timer will start. If this user is classified as still, the timer value will increase for the time the user is in this state. If the user moves, the timer will restart. If the timer exceeds a time threshold defined in system configuration, it will send an alarm signal to the main control (system configuration allows the use of different time threshold for different room zones).
Furthermore, and although this system works processing video sequences, these videos are not stored in local memory or external memory as a server or a Cloud platform. The algorithms process the images online and only send the necessary alarms to the main control when an anomaly is detected. Therefore, people privacy is guaranteed in all moment.
This system has been developed sharing the same communication architecture developed by Ibermatica in their smart sensors so it can be added as a new node to the home monitoring sensor net.