Wearing-independent hand gesture recognition method based on EMG armband

Abstract

Electromyographic (EMG) armband with electrodes mounted around the user's forearm is one of the most ergonomic wearable EMG devices and is used to recognize fine hand gesture with great popularity. Definitely, the distributions of signal differ greatly in different wearing positions of armband based on the physiological characters of EMG, which will cause the performance decline and even the inapplicability of the recognition model built in one position. Hence, this paper proposes a wearing-independent hand gesture recognition method based on EMG armband. To eliminate the influence of wearing position, Standard Space is proposed in this paper. Based on the sequential features of EMG in different scales, the wearing position of armband is predicted and helps unify the original features to the proposed space. Then, with the unified signals, fine hand gesture can be recognized accurately and robustly with lightweight Random Forest (RF). The experimental results showed that the recognition accuracy of the proposed method was 91.47% approximately. And compared with the method without fine feature extraction and feature space unification, the performance was improved by 10.12%.