Physical activity classification using a single triaxial accelerometer based on HMM

This study focuses on physical activity classification method using a single triaxial accelerometer attached on chest. With acceleration data acquired by a wearable wireless device, features are extracted using sliding window to describe different activity types. Hidden Markov Model (HMM) is used to recognize physical activity sequence. A modified Viterbi algorithm is used to find the optimal state sequence. The experimental results on 6 subjects have achieved an overall accuracy of 99.59% using our method, which is the best result so far.