sEMG-Based Small Rotation Invariant Gesture Recognition Using Multivariate Fast Iterative Filtering

Surface electromyography (sEMG) is an important tool for pattern recognition in modern society. Electrode shift is a major challenge in sEMG-based systems and affects the performance greatly. In this letter, a method is suggested for hand gesture recognition using sEMG, which is suitable for small angle electrode rotation scenario. A root-mean-square-based envelope is employed for segmentation followed by sEMG signals decomposition using multivariate fast iterative filtering. Moreover, time domain-based features are computed and given to the classification model. The classification model is trained with the initial position of sEMG electrodes and tested with small angle rotations i.e., 0 $^{\circ }$ , 10 $^{\circ }$ , 350 $^{\circ }$ , 20 $^{\circ }$ , and 340 $^{\circ }$ . Efficacy of the designed method is investigated against eight different hand gestures. The suggested method achieved 88.82%, 82.54%, 76.98%, 68.25%, and 61.11% accuracy in case of 0 $^{\circ }$ , 10 $^{\circ }$ , 350 $^{\circ }$ , 20 $^{\circ }$ , and 340 $^{\circ }$ sEMG electrode shift, respectively, and outperforms the compared method.