Towards Hand-Object Gesture Extraction from Depth Image

Abstract

Electromyographic (EMG) signal is the electrical manifestation of a muscle contraction. Surface EMG signal can be obtained by electrodes on the skin to control prosthetic hand. However, surface EMG is sensitive to environmental interference, which leads to a low motion recognition rate of prosthesis control when encountering unexpected interferences, like electrodes shift. Electrodes shift occurs particularly in the day-today use of wearing electrodes. As a reslut, a long-term training procedure is necessary. To solve this problem, this paper proposes a new sEMG electrodes configuration to reduce the interference caused by electrodes shift. Experiments are designed to verify the improvements through evaluating the classification accuracy of discriminating eleven hand motions by pattern recognition approach. The comparison results show that the proposed electrodes configuration increases the pattern recognition rate by 4% and 8% when applied kNN and LDA classifier, respectively. This paper suggests that optimising electrodes configuration is able to improve the EMG pattern discrimination and the proposed electrodes configuration has reference value.