Automatic EMG-based Hand Gesture Recognition System using Time-Domain Descriptors and Fully-Connected Neural Networks

Abstract

Hand gesture recognition has numerous applications in medical (e.g., prosthetics), engineering (e.g., robot manipulation) and, even, military research areas (e.g., UAV control applications). This paper proposes a fast and accurate method to identify hand gesture categories based on electromyo-graphic (EMG) signals registered by a commercial sensor (e.g., Myo Armband developed by Ontario-based Thalmic Labs), which is placed on the user's forearm. The proposed method is based on the extraction of time-domain features and a neural network architecture to perform the classification of the EMG signals. In order to evaluate the performance of the proposed algorithm, we use a publicly available dataset with 7 hand gesture categories. The proposed hand gesture recognition system achieves a 99.78 % overall performance accuracy, which is comparable to that reported by applying other state-of-the-art methods, but is able to work in real-time conditions.