Introducing a Novel sEMG ANN-Based Regression Approach for Elbow Motion Interpolation

Abstract

Surface electromyogram (sEMG) signals are extensively used for rehabilitation and control purposes. However due to their intrinsic complexities and intense sensor crosstalk, feature classification and pattern recognition of sEMG signals especially for motion analysis are quite challenging. This study proposes a versatile sEMG Artificial Neural Network based regression approach to evaluate a simple elbow motion with respect to a reference frame. The proposed approach attempts to appropriately interpolate intermediate position angles in an attempt to evaluate and substantiate a continuous motion of the forearm. Results show that based on the proposed algorithm, with a correlation of about 91% for the test data, it is possible to track the motion of the forearm through a set of discrete sEMG signals.