Intelligent Recognition of Upper-Limb Muscle Activation Based on Surface Electromyography and Kinematic Information Fusion

Abstract

To address the difficulty of accurately identifying active muscle activation during upper-limb rehabilitation after stroke, this study proposes a muscle synergy recognition method based on surface electromyography, kinematic parameters, and game-task data. Surface electromyographic signals are collected from five upper-limb muscles, including the anterior deltoid, posterior deltoid, biceps brachii, triceps brachii, and brachioradialis. Meanwhile, shoulder and elbow quaternion data are obtained using a wearable motion-capture device to estimate elbow flexion and extension angles, and task performance data are recorded during rehabilitation games. After preprocessing and feature extraction, a neural network model is constructed to identify muscle activation patterns in different movement phases. Furthermore, the mapping relationship among sEMG features, elbow motion angles, and game performance is analyzed. This study provides a quantitative basis for recognizing upper-limb muscle coordination and supports future active intention recognition and adaptive assistance control in rehabilitation robots.