Decoding surface electromyogram into dynamic state to extract dynamic motor control strategy of human

We propose a method to decode surface electromyogram (sEMG) data into the dynamic state that delivers the characteristics of dynamic motor control strategy of humans. First we propose the clustering and the segmentation of the dynamic state by using a Bayesian mixture of Gaussian model (Bayesian MoG) in the augmented space of both hand position and sEMG. Second, we introduce a hidden semi-Markov model (HSMM) to decode sEMG into the dynamic state similar as much as with the segmented result without hand position. Experimental data were collected to train both Bayesian MoG and HSMM, and cross-validation between segmentation and the decoding result were performed to verify the decoding accuracy of the HSMM. Finally, we verified that the decoding results successfully extracted a dynamic motor control strategy from sEMG data.