Hand Motion Classification Using sEMG Signals Recorded from Dry and Wet Electrodes with Machine Learning

Abstract

Commercial biopotential electrodes for Electromyography (EMG), Electrocardiography (ECG) and Electroencephalography (EEG) are rigid and inflexible. Due to the use of conductive gels and the adhesive pads, these electrodes are disposable, thus unsuitable for long-term use. Flexible dry electrodes in the literature have been reported to record the biological signals relatively well. Although there are still challenges in this emerging field of research to overcome in the near future. In this manuscript, we have fabricated flexible dry electrodes to record the surface EMG (sEMG) signals. Five hand motions are classified based on the sEMG signals recorded by our electrodes as well as commercially available wet rigid electrodes. The recognition accuracy reaches proximately 85% and 82% for the sEMG signals recorded by the dry and wet electrodes, respectively. Gaussian SVM algorithm is used as the most appropriate classifier here. The classifier accuracy is found out to be relatively similar comparing the commercial wet electrodes to our fabricated dry counterparts.