Real-time Physical Activity classification and tracking using wearble sensors

A sensor-network physical activity monitoring system (SAMS) using wearable sensors is presented. It classifies and tracks body activity in real time. The system adopts a service-oriented architecture for data acquisition, sensor actuation & control, and real-time service management. The activity recognition and tracking are carried out using two levels of approaches. At level 1, the movement and flexion angles of body segments are recovered from accelerometer signal by using extended Kalman filter (EKF). At level 2, the movements of body segments are further aggregated by using Hidden Markov Model (HMM) for activity recognition. The proposed system is applied to monitoring and identifying normal daily activities in an apartment. Experimental results indicate that, by using the proposed system, body activity can be identified with high accuracy and short system latency.