EMG Based Gesture Recognition by Visual Method for Human-Machine Interaction and System Health Management

Human-machine interaction refers to the communication and interaction between human and machine through a user interface. Surface electromyography-based gesture recognition, as a non-invasive human-machine interaction method, has been widely used in medical, entertainment and industrial fields. This paper proposes a framework to recognize EMG signal gesture through deep learning visual method for human-machine interaction. We introduce two tricks into EMG signal gesture recognition: I) fully convolutional network, II) shortcut connection. An abundant experiment was conducted with a total 49 of gestures on Ninaweb DB2 and an average accuracy of 64.02% was achieved. This result indicates that the proposed method provides an effective approach to gesture recognition for human-machine interaction and system health management.