GenEMG-Rehab: AI-Powered Wearable Surface EMG System for Intelligent Gesture Recognition and Early Muscle Weakness Assessment

Abstract

Surface electromyography (sEMG) is a technology that has been developed which is non-invasive and can monitor muscle activity, and interaction with a machine. In this paper, we present a clever system for hand gesture recognition and assessment of early muscle weakness based on Artificial Intelligence (AI) and wearable Surface Electromyography (EMG). The proposed system is planned to acquire the muscle sEMG signals from the muscles of the forearm surface electrodes set on the forearm muscle while the forearm muscles which are to be measured are around the forearm. Acquired signals are subjected to signal conditioning that includes the processing of amplification, filtering, rectifying and normalizing the signal. The Root Mean Square (RMS), Mean Absolute Value (MAV), Waveform Length (WL) and Zero Crossings (ZC) are Root Mean Square extracted time domain features, used for classification. Following the proposed hand gesture recognition system is a hybrid convolutional neural network-long shortterm memory (CNN-LSTM) model for hand gesture recognition along with activation patterns of the muscles. The system is able to detect muscle fatigue and give feedback for rehabilitation, besides gesture recognition. Experimental evaluation indicates that the gesture classification and monitoring of the muscle activity are correct and that the technology works well. The proposed framework can find applications in rehabilitation engineering, control of prostheses, assistive technologies, telemedicine and smart healthcare system.