Video Based Fall Detection with Enhanced Motion History Images

Abstract

Computer vision systems offer a new promising solution which can help older people stay at home by providing a secure environment and improve their quality of life. One application area of video surveillance is to analyse human behaviour and detect unusual behaviour. Falls are one of the greatest risks for the elderly living at home. This paper presents a novel approach for detecting falls, based on a combination of motion information and human shape variation. The motion information of a segmented silhouette, when extracted can provide a useful cue for classifying different behaviours. Also, the variation in human shape can used to establish the pose and hence fall events. The approach presented here extracts motion information, use variation in shape and in addition use best-fit approximated ellipse around the human body to further improved the accuracy of falls detection. Result of our approach demonstrates a 20% improvement over motion information only implementations.