An SVM classifier for detecting merged motor unit potential trains extracted by EMG signal decomposition using their MUP shape information

Detecting merged motor unit potential trains (MUPTs) during electromyographic (EMG) signal decomposition can assist with improving decomposition results. In addition, such invalid MUPTs must be identified and then either corrected or excluded before extracted MUPTs are quantitatively analyzed. In this work, a support vector machine (SVM) based supervised classifier that evaluates the shapes of the motor unit potentials (MUPs) of a MUPT to determine whether it represents a single MU (i.e. it is a single MUPT) or not, is described. A given MUPT is represented by six MUP-shape based features and then assessed using the SVM classifier. Evaluations performed using several simulated EMG signals show that the SVM, with overall accuracy of 95.6%, performed significantly better than Fisher discriminant analysis and logistic regression based classifiers. The SVM correctly classified 98.5% of the single trains and 79% of the merged trains.