Using the Robust High Density-surface Electromyography Features for Real-Time Hand Gestures Classification

Using High-Density surface Electromyography (HD-sEMG) signals for gesture classification has augmented the spatial information of muscle activity by increasing the density and convergence of the electrodes. In this paper, spatial features are extracted from HD-sEMG data. These features generated by combining HOG features of HD-sEMG map and intensity features calculated from the average of segmented HD-sEMG map which is denoted as (AIH) features. Real-time evaluation was performed for inter-session identification. The simulation of proposed algorithms is achieved by MATLAB; the result of our experiments achieves high accuracy with good performance based on spatial features reached to 99%. The comparison of our results with other research indicates that the proposed algorithms can enhance the performance and accuracy of gestures identification process by SVM classifier. In addition, the results confirm the robustness of the spatial features to the variation of EMG signals over time.