Hand gesture and activity recognition in assisted living through wearable sensing and computing

Abstract

With the growth of the elderly population, more seniors live alone as sole occupants of a private dwelling than any other population groups. Helping them to live a better life is very important and has great societal benefits. Assisted living systems can provide support to elderly people in their houses or apartments. Since automated recognition of human gestures and activities is indispensable for human-robot interaction (HRI) in assisted living systems, this dissertation focuses on developing a theoretical framework for human gesture, daily activity recognition and anomaly detection. First, we introduce two prototypes of wearable sensors for motion data collection used in this project. Second, gesture recognition algorithms are developed to recognize explicit human intention. Third, body activity recognition algorithms are presented with different sensor setups. Fourth, complex daily activities, which consist of body activities and hand gestures simultaneously, are recognized using a dynamic Bayesian network (DBN). Fifth, a coherent anomaly detection framework is built to detect four types of abnormal behaviors in human's daily life. Our work can be extended in several directions in the future.