University of AlbertaMultimedia Research Centre · Dept. of Computing Science
ROSSRemote Observation, Sensing & System
Research / R/05 Health & Tele-medicine / Monitoring health and safety at a distance
R/05 · Health & Tele-medicine

Monitoring health and safety at a distance

In a long-term care room, a camera can watch for falls without ever storing a recognisable image. The group extracts skeletons and vital signs, processes them locally, and raises an alert.

Health & Tele-medicineFall DetectionPrivacyEldercare

Some of the most valuable remote observation happens indoors, at arm’s length. In a long-term care facility, continuous monitoring can catch a fall or a change in vital signs long before the next staff round. The challenge is doing it without turning a resident’s room into a surveillance camera.

Privacy is the design constraint, not an afterthought. Rather than store video, the group extracts a skeleton: the joints and limbs of a body reduced to moving points. A fall is clear in the skeleton’s motion, but the person is not identifiable from it.

A body reduced to a skeleton of joints for privacy-preserving fall detection

Joint and skeleton extraction. Movement and posture are preserved; identity is not.

Everything runs on the edge. In a fielded Mitacs project, the video is processed locally inside a secured, locked box in the room, so raw imagery never leaves the premises. Only the result (a fall event or alert) is sent onward to a mobile app for a caregiver.

The same signals carry health information beyond safety. Contact-free sensing yields vital-sign traces that can be tracked over time, for several residents at once.

Vital-sign signals tracked over time for several monitored subjects

Vital-sign traces recorded over time. Long-baseline signals like these turn a single alert into a picture of trend and change.

Watching for a fall, reading a pulse, raising an alert: it is the same acquire, analyse, transmit loop as satellite monitoring, moved indoors and built to protect the person it watches.