An Edge-AI Pose-Estimation Platform for Real-Time Posture and Posture-Transition Monitoring: A Feasibility Study for Pre-Emptive Fall-Risk Alerting in Elderly and Dementia Care

A real-time Edge-AI posture monitoring platform using 17-keypoint pose estimation classifies lying, Fowlers, sitting, and standing at 30 frames per second on a Raspberry Pi 5 with Hailo-8L.

• Why it matters: Falls and unmonitored movements in elderly and dementia care are often addressed reactively with labor-intensive or false-alarm-prone approaches, creating care and staff-efficiency gaps.
• What they did: The system runs Hailo-YOLOv8-Pose locally on a Raspberry Pi 5 with a Raspberry Pi camera module 3, transmitting alerts via Bluetooth to caregivers while evaluating bed/chair monitoring, occlusion, free limb movement, and varying lighting.
• The result: It maintained accurate monitoring under varying lighting, including lights-off scenarios, and preserves privacy by using skeletal keypoints without recording video, enabling proactive fall-risk alerting and resilience during network outages.

AI summary of the abstract · may contain errors · not medical advice · read the paper.