Upper-limb movement classification through logistic regression sEMG signal processing

Abstract

Several computational intelligence algorithms have been used to classify biological signals of stochastic nature. This paper aims to evaluate the application of logistic regression technique for the classification of electromyography signals originated from hand-arm segment. Therefore, the algorithm was implemented using multinomial logistic regression and an optimization heuristic based on gradient descent. Classification tests were performed with three subjects and an accuracy rate of 90.2 ± 3.8% was achieved.