Information Based Classification Of Motor Unit Action Potentials

The decomposition of myoelectric (ME) signals requires the detection and classification of motor unit action potentials (MUAPs). This paper concentrates on the classification of MUAPs. The intent was to compare the performance of a probabilistic inference approach to the classification of MUAPs with traditional template matching schemes. A simulation program was used to generate ME signals, from which the analyzed MUAPs were extracted. The comparison of the two classification techniques was based solely on time samples. It has been found that the probabilistic inference technique can achieve a performance level comparable to the template matching method. It has also been suggested that the full potential of the probabilistic inference technique has not been demonstrated under the conditions of this study. Should this potential be realised, the technique presented here could well outperform the traditional template matching methods.