Using AI to protect the elderly

UAlberta computing scientists develop autonomous intelligence system to improve independence and safety for seniors.

Katie Willis - 24 October 2018

An autonomous intelligence system is helping seniors stay safe both at home and in care facilities, thanks to a collaboration between University of Alberta computing scientists and software technology company Spxtrm AI.

The new tool uses a deep-learning computer vision system and motion-classification algorithms to capture events such as falls in real time, alert caregivers, and give health-care professionals the information they need for immediate triage. The system was developed in part by the Multimedia Research Centre, led by Irene Cheng in the Department of Computing Science.

"Just-in-time action is needed for falls and other accidents in order to save lives, and only accurate and time-efficient algorithms can deliver real-time solutions," explained Cheng.
"Videos are captured continuously, at all hours, and at high resolution. It is impossible for humans to monitor these systems and detect the relevant information in real time as effectively as this autonomous system can."

The system transfers real-time video to an autonomous computer vision lockbox. If an event, such as a fall, is detected, the system alerts a specified caregiver and provides a redacted video of the event. The system maintains the privacy of the user, while providing the caregiver with important triage information, including the moment of impact after the fall.

"Privacy is a major concern for most seniors," said Cheng. "Our algorithms are able to extract the necessary information on the fall for analysis without disclosing their physical appearance to human operators and caregivers."

The research was funded by a multi-year Mitacs grant, supporting graduate students and researchers in their research and development of the project.

"We are very pleased to be working with the University of Alberta as it provides us with a team of bright and talented machine learning researchers and developers that supplements our own machine learning team as well as providing us amazing candidates for future full-time positions," said Jesse Slade Shantz, Spxtrm AI's chief medical officer and co-founder.