AICARE: Health Abnormality Detection of Elderly Automatically Using Deep Learning

Elderly people who are far from relatives or live alone need regular attention to timely detect abnormalities in their health. To solve this problem, technology has been applied to make it easier and more convenient to monitor the health of the elderly remotely. In this work, we proposed a method and system for automatically tracking and detecting behavioral abnormalities of the elderly via surveillance cameras using deep learning. In our methodology, the elderly with various postures such as standing, sitting, lying, chest holding will be detected and tracked when appearing on surveillance cameras in their homes. From postures identified in continuous frames during the tracking process, behavioral abnormalities can be detected and can be seen as signs of health problems such as fall, faint, stroke, heart attack. We experimented on our dataset collected from private surveillance cameras with normal quality. The result showed that our methodology can be used with cheap cameras to provide service for massive homes on elderly health abnormalities tracking and early warning.