In care facilities, falls are a frequent and serious risk — particularly for elderly residents. Staff cannot monitor all areas continuously, and existing call-button systems rely on the resident being able to press a button after a fall.
We developed a camera-based fall detection system that uses pose estimation and motion analysis to identify fall events in common areas and corridors. When a fall is detected, the system immediately alerts on-duty staff with the location and a snapshot for verification.
Privacy was a core design requirement. The system processes video locally and does not store raw footage — only anonymized event data and alert records are retained. The detection model distinguishes between falls and normal low-posture activities (bending, sitting down) to minimize false alarms.
The system operates as an additional safety layer alongside existing care protocols, helping staff respond faster to incidents that would otherwise go unnoticed until the next scheduled check.