Depth camera based fall detection using human shape and movement

The number of elderly people living alone have increased over the last years and fall is one of major risks that threaten their lives. Computer vision is one of the accurate solution for fall detection. In this paper, we propose a new method for fall detection using depth camera. This method combines human shape analysis, head tracking and center of mass detection by exploiting the advantages of Kinect. In addition, we take into account the motion information, and use the relationship between time and distance translated by covariance to discriminate falls. The experiments with SDUFall dataset which contains 20 subjects performing five daily activities and falls demonstrate that the proposed method can achieve up to 92.98% accuracy.