Wavelet analysis for Support Vector Machine classification of motor unit action potentials

The paper presents a new method for neuromuscular disorders diagnosis based on analysis of scalograms determined by the Symlet 4 wavelets technique. Obtained results served for extraction of five features, which, after SVM analysis, were reduced to a single decision parameter allowing assigning the investigated cases to one of three groups: myogenic, neurogenic or normal. Software implementation of the method permitted to create a diagnostic tool for EMG investigation aid. The method characterizes high probability of accurate diagnosis of a muscle state with total error of 0.5% - 4 misclassifications out of 780 examined cases.