Real-Time Hand Gesture Recognition Model Using Deep Learning Techniques and EMG Signals

摘要

Gesture recognition has multiple applications in medicine, engineering and robotics. It also allows us to develop new and more natural approaches to human-machine interaction. Real-time hand gesture recognition consists of identifying, with no perceivable delay, a given gesture performed by the hand at any moment. In this paper, we propose a model for real-time hand gesture recognition. The proposed model takes as input electromyographic (EMG) signals measured on the forearm, using the commercial sensor Myo Armband. We use an autoencoder for automatic feature extraction, and an artificial feed-forward neural network for classification. The proposed model can recognize the same 5 gestures as the proprietary recognition system of the Myo Armband, achieving an average recognition accuracy of 85.08% ± 15.21%, with an average response time of 3 ± 1 ms. The proposed model is general, which implies that it can recognize the gestures from any user, even when his\her data are not included in the training dataset. Finally, for reproducing this work, we make publicly available the code that implements the proposed model.