Automated fall detection using computer vision

Abstract

The population of elderly people is increasing day-by-day in the world. One of the major health issues of an old person is injury during a fall and this issue becomes compounded for elderly people living alone. In this paper, we propose a novel framework for automated fall detection of a person from videos. Background subtraction is used to detect the moving person in the video. Different features are extracted by applying rectangle and ellipse on human shape to detect the fall of a person. Experiments have been carried out on the UR Fall Dataset which is publicly available. The proposed method is compared with existing methods and significantly better results are achieved.