Paul Etienne

Seculity: fall and safety monitoring for older adults, without cameras watching them Problem: Older adults living alone need someone to know when they fall or stop moving. Camera monitoring feels invasive, and wearable alarms only work if the person remembers to wear them and can press the button. Solution: Seculity turns an ordinary home camera into a privacy-first monitor. Computer vision reads only body posture. Families see a 3D avatar in a virtual garden that mirrors it (standing, sitting or lying down), never video. Meta’s Muse Spark turns the posture timeline into the words a caregiver needs: instant “possible fall” notifications and plain-language daily updates. How it works: YOLOv8 and MediaPipe Pose run on a single CPU core and extract body joints. A posture classifier and fall rules turn those into events. The events, never images, go to Muse Spark on the Meta Model API, which writes the alerts and summaries. A three.js viewer shows the avatar and the caregiver panel. Results: In a working proof of concept, the avatar follows the person about one second behind real life, at 11 to 15 fps. Standing, sitting and lying are recognized correctly in daylight and night vision. A simulated fall triggered an alert instantly, with Muse Spark’s wording about 11 seconds later. Next: Move processing onto an in-home hub so video never leaves the house, and train a model that recognizes falls from short sequences of movement. Also add support for multiple residents.