MUAP Classification Based on Wavelet Packet and Fuzzy Clustering Technique

Abstract

The present study compares several methods with regard to feature extraction and classification of motor unit action potentials (MUAPs) for electromyography (EMG) signal decomposition. The technique was applied to single-channel, short-period real myoelectric signals from normal subjects and artificially generated EMG recordings. All the real EMG recordings were made from the biceps brachii of healthy subjects during voluntary contraction at different force. A model, based on the phenomenon of EMG signal, is used to test the proposed technique on synthetic signals with known features. In contrast to previously developed methods based on EMG signal decomposition performance, our technique has two important distinctive characteristics. Firstly, we applied the local discriminant optimal wavelet packet for the feature extraction of MUAPs. Secondly, we optimized the MUAP classification result using the fuzzy C-means clustering technique to improve the EMG decomposition accuracy. Therefore, the method is substantially automatic and has been evaluated with synthetic and experimentally recorded myoelectric signals.