The Independent Component Approach to the Surface Multi-channel EMGs Decomposition Method

Introduction It is electromyogram (EMG) that represents interference signals of electrical activities in many motor units (MUs). An EMG decomposition method is the procedure by which the signal is classified into its constituent motor unit action potential trains (MUAPTs). So far the classification to each MUAPT in multi-channel EMGs by means of an EMG decomposition method is directly performed with differences of the waveform’s characteristics in each MUAPT (Mambrito and DeLuca, 1984). The presently employed EMG decomposition methods have the following two problems: (1) only records derived from attempted isometric contractions have been decomposed because most of EMG decomposition methods employ invasive needle EMGs; (2) the processing spends extremely enormous time because it is an interactive method between an operator and the system. On the other hand, the use of multi-channel surface EMGs has such a high applicability to measure a variety of neuromuscular activities, e.g., position of innervation zone, action potential conduction velocity and identification of a single MU. Recently, independent component analysis (ICA) has been shown to be an efficient tool for feature extraction and blind source separation (BBS) from EEG and MEG and other multi-channel bio-signal recordings (Vigario et al. 2000, Lathauwer et al., 2000). ICA maximizes the forth order statistics, called by “kurtosis”, to maximize the statistical independence between the channels. By using ICA, multi-channel signals are extracted to each subspace. The purpose of this study is to examine the emerging technique of ICA for EMG decomposition by means of the simulated multi-channel surface EMGs. In this paper, we organized the EMG model by means of a traveling dipole (Fuglevand et al., 1992) and discuss the results performed on the simulated multi-channel surface EMGs with ICA.