Decision making in electromyography using wavelet-type analysis and fuzzy clustering

Classification of motor unit action potentials in electromyography is to be based on an optimal method for feature extraction, matched to the special characteristics of the signal, and on an efficient method of pattern analysis. For the feature extraction stage, wavelet-type representation of the motor unit action potentials has been compared to conventional orthogonal decomposition using Karhunen-Loewe transformation (KLT). Classification of the feature vectors was carried out using a modified version of the unsupervised optimal fuzzy clustering algorithm (UOFC). By application of the algorithms to test data comprised of 130 labeled motor unit action potentials it could be verified that the wavelet-type decomposition was significantly superior to the KLT