Surface EMG decomposition based on innervation zone mapping and an LMMSE framework

Abstract

Muscle motor units (MUs) can provide valuable information on neuromuscular control and muscle diseases, and electromyogram (EMG) decomposition is an effective method for reconstructing MUs. In this paper, a novel decomposition method based on innervation zone mapping (IZM) and a linear minimum mean square error estimation (LMMSE) framework is proposed for high-density surface EMG (HD-sEMG) decomposition. First, initial discharges were selected according to the IZM at each discharge. Then, the LMMSE framework and multistep iterative process were employed to update and estimate the innervation pulse train (IPT). Each discharge in the IPT was then classified into an individual MU action potential train (MUAPT) according to the IZM at each discharge. Finally, the MUAPTs with the same IZMs at the initial discharge were merged. The method based on the IZM and the LMMSE framework (IZM-LMMSE) was validated on both simulated and experimental data. By comparing the decomposition results of IZM-LMMSE and K-means clustering-modified CKC (KmCKC), the IZM-LMMSE algorithm can obtain more MUs and higher accuracy in most cases. Moreover, the experimental results show that the ratio of all common discharges obtained by IZM-LMMSE and KmCKC is 0.94 ± 0.01 % (mean ± std), which proves the effectiveness of the IZM-LMMSE method. The decomposition method has wide application prospects in rehabilitation engineering and clinical diagnosis.