Automated abnormal behavior detection for ubiquitous healthcare application in daytime and nighttime

Abnormal behaviors such as falls are the most significant issues in ubiquitous healthcare applications for the elderly. The goal of this research is to develop a novel computer vision algorithm, which can allow discriminating between abnormal behaviors and normal daily activities using statistical moment analysis and human shape analysis in daytime and nighttime environment. Many researchers recently presented their work aimed at developing an image sensor based application to detect unexpected fall events. Until now, most studies on the abnormal behavior detection of the elderly have been presented for the systems in daytime. However, they cannot give any solution to detect falls in both daytime and nighttime. To overcome the problem, our method is implemented to distinguish abnormal activities from normal activities. Experimental results show very promising results on test image sequences of normal daily activities and simulated fall activities.