Bone density, strength, and coordination all decrease as we age, and this fact can lead to some serious consequences in the form of slips, falls, and other accidents. In Finland, falling is the most common type of accidental death among those age 65 and over, amounting to around 1,200 per year. But Thomas Vikstrom hopes to decrease this number by detecting falls the moment they occur through the use of the Arduino Nicla Sense ME’s accelerometer together with a K-Way jacket and a smartwatch.

At first, Vikstrom tried to gather and label data for all kinds of activities, including sitting, walking, running, driving, etc., but later realized anomaly detection would be much better suited for this application. After collecting around 80 seconds of data with Edge Impulse Studio, he trained an anomaly detection model to detect when any out-of-the-ordinary events occur. The model was then deployed to the Nicla Sense ME by integrating the inferencing code with a BLE service that outputs a positive value when a fall is detected, as well as illuminating the onboard LED.

To receive this information, Vikstrom added a Bangle.js 2 smartwatch to the system which automatically calls an emergency number if the wearer fails to intervene. For more details, you can check out his Edge Impulse docs page here. Although only a proof of concept, this K-Way project demonstrates how tinyML-powered outerwear can be used to detect falls, and together with cellular network devices send for help in case the user is immobile.