Advanced Signal Processing Techniques for Multi-channel EMG – On the Need for Motor Unit Action Potential Compensation

Abstract

We systematically assessed the impact of motor unit action potential (MUAP) compensation on muscle excitation estimation from high-density EMG (hdEMG). For this purpose, we used experimentally recorded hdEMG signals from biceps brachii to estimate the MUAPs of 200 MUs at different elbow angles. We then used these MUAPs to generate synthetic EMG signals with known muscle shortening and muscle excitation profiles. Both constant and force varying isometric muscle contractions were analyzed. Novel metric for muscle excitation estimation, so called global activity index (GAI) was introduced and compared to cumulative motor unit spike train (CST), identified by Convolution Kernel Compensation (CKC) method and to spatial average of root-mean-square (RMS) value of hdEMG signals. The results of all three metrics were compared to the CST of all the simulated motor units by calculating normalized RMS error (NRMSE). Processing costs of investigated methods were assessed by measuring the processing time on a personal computer with CORE i7 processor. The results demonstrate that the GAI significantly outperforms the RMS metric and is comparable to the CKC CST at significantly lower computational costs.