Finger motion recognition robust to diverse arm postures using EMG and accelerometer

Abstract

The electromyogram (EMG) based finger motion recognition accuracy may be degraded during the actual stage of practical applications due to various causes. Among them, the representative issue is the changes of the EMG signals of the identical finger motion by the different arm postures. We propose an EMG based finger motion recognition technique robust to diverse arm postures. The proposed method uses both the signals of the accelerometer and EMG simultaneously to recognize correct finger motions for each arm posture. We compared the experimental results with and without considering the corresponding arm postures to recognize finger motions. The average recognition of finger motions with the correct arm posture inference was 85.7% which is 31.6% higher than without considering the corresponding arm postures. In this study, accelerometer and EMG signals were used simultaneously, which decreased the effect of different arm postures on the EMG signals and therefore improved the recognition accuracy of finger motions.