Real-time physical activity monitoring by data fusion in body sensor networks

Abstract

A physical activity monitoring system by data fusion in body sensor networks is presented in this paper, which targets at providing body status information in real time and identifying body activities. By fusion of data collected from several accelerometer sensors placed on different parts of the body, the activities can be identified and tracked Mathematical approaches employed in the system include Kalman filter and hidden Markov model. With the proposed system architecture, these algorithms are distributed to different components of the system. The proposed system is applied to monitoring and identifying daily activities in laboratory and comparatively intensive activities in a gym room. Video-based approach is used as the benchmark to evaluate its performance. Comparative results indicate that, by using the proposed system, body status of daily activities can be estimated with good accuracy in real time, and body activity is identified with high accuracy within short system latency.