sEMG pattern classification using hierarchical Bayesian model

This work addresses surface electromyogram (sEMG)-based muscle pattern classification using a generative model. By using a hierarchical Bayesian model, the proposed approach constructs an overall process model of recorded sEMG signals. By inferring probabilistically latent neural states which governs a collection of training sEMG data, classification is realized. To validate the approach, eight-class classification using four sEMG sensors on the limb actions is tested with five subjects. The proposed model achieves an overall 95% accuracy in the classification experiment. The results support that the proposed approach is very promising for sEMG pattern classification.