Automatic classification of electromyographic signals

Abstract The results of the application of classification methods to electromyograph signals of weak contractions in normal and myopathic subjects are described. Methods of pattern recognition, previously presented, allow the selection of representative motor unit action potentials. The analysis is done with ordinal qualitative variables obtained by identification of shape descriptive parameters (amplitude, duration, number of phases, number of extrema). From this analysis, characteristic classes for normality and myopathy appear, from which a diagnostic aid by assignment can be made.