Ambient Assisted Living IoT Solutions === ## Background ### Use cases The aim is to find an IoT device which provides the means to alert a carer, monitoring a dementia patient living independently, to the following events : * Falls * Wandering * Appliances left on * Fire ### Prerequisites and Constraints It is desirable to have an option which is "commercial off-the-shelf" (COTS) to avoid hardware development, monolithic to reduce overall complexity and cost, and wall-mounted to retain patient dignity and reduce the chances of being forgotten or misplaced. Furthermore, the device must be discreet so as not to impinge upon a patient's privacy or dignity. The device would ideally be battery powered with a long service interval, independent of WiFi or cellular infrastructure, and would provide data via an easily accessible cloud API to facilitate integration into existing software solutions. Camera based solutions are prohibitive due to the potential to breach GDPR regulations and development of the computer vision algorithms required to enable the above use cases may well take more time than the pre-accelerator programme can offer. ### Required Sensors To reduce overall system complexity, ideally the number of sensors in the solution should be kept to a minimum. Therefore, we must identify what is the minimum set of sensors capable of directly or indirectly inferring the events listed above. #### Falls * Noise level (sound of impact, shouting) * Motion (motion detected towards floor, followed by lack of motion) * Thermal imaging (no motion, but heat signature indicates person staying still) * Accelerometer (vibration/bang caused by impact) #### Wandering * Thermometer (door opening cause sudden change in internal temperature) * Light level (sudden change in light level indicates door was opened - may be unreliable depending on time of day) * Motion (Lack of motion in all sensors shows no occupancy) * Thermal imaging (Combined with motion above, confirms no occupancy) #### Appliances Left On A combination the following can potentially be used to infer that an appliance was left on for an extended period with no-one in the vicinity. * Sound * Light Level * Thermal Imaging * Motion #### Fire A combination of the following can be used to infer a fire is present, for obvious reasons hopefully. * Sound (Shouting) * Light Level * Temperature * Smoke/particulate detector * Thermal Imaging #### Summary Overall, a general-purpose monolithic sensor unit **must** implement the following sensors to cover the use cases above: * Thermal imaging (low resolution) * Motion (PIR sensor) * Sound (min/mean/max decibel level) * Light level (Luminence Sensor) and to provide supplementary functionality, the unit **could** implement the following sensors: * Accelerometer (vibration detection) * Gyroscope/Magnetometer (sensor fusion with above to determine sensor orientation) * Smoke/Particulate sensor (fire and air quality warnings) ## Market Research ### COTS Wearable (Undesirable) * [Care band](https://www.carebandremembers.com/careband-health-safety/dementia-alzheimers/) ### COTS Wall Mounted * [Elsys Eye](https://www.elsys.se/en/ers-eye/) (£85-116) * [Tektelic Smart Room Sensor](https://tektelic.com/catalog/smart-room-lorawan-sensor-pir) (£50) ### DIY Wall Mounted * [Grid Eye Thermal Sensor](https://shop.pimoroni.com/products/adafruit-amg8833-ir-thermal-camera-breakout?variant=984585994250) (£40) * Option 1 (Easier but no battery option) * Raspberry Pi Zero 1/2 W (£5 - £15) * [Enviro Hat](https://shop.pimoroni.com/products/enviro?variant=31155658489939) (£19-36) * [LoRa Hat](https://uk.pi-supply.com/products/iot-lora-node-phat-for-raspberry-pi) (£30) * Option 2 (Fully customizable - battery possible) * Raspberry Pi Pico (£4) * Environmental sensors * Accelerometer/Gyroscope * Light level * Microphone ## Development Notes ### Thermal Imaging Person detection demo: https://www.hackster.io/naveenbskumar/intruder-detection-using-ultra-low-powered-thermal-vision-24ed45#toc-hardware-setup-4 Interpolation: https://learn.adafruit.com/adafruit-amg8833-8x8-thermal-camera-sensor/arduino-wiring-test?view=all