An Elbow Bilateral Rehabilitation System Based on Surface Electromyogram: Design and Validation

Recent years have witnessed the promising future of surface electromyography (sEMG)-based motion intention recognition technology in the intelligent control of rehabilitation exoskeletons. In this paper, we propose an elbow bilateral rehabilitation system (EBRS) based on sEMG, which achieves continuous tracking of the healthy limb movement intention by the affected limb. We develop the hardware and control system for the EBRS. Additionally, we explore various combinations of nine sEMG features with three regression algorithms to identify the optimal combination for estimating elbow joint angles. Experimental results with subjects showed that utilizing the Generalized Regression Neural Network (GRNN) in combination with Root Mean Square (RMS) and Waveform Length (WL) features yielded the second-bast regression performance (RMSE: 0.903; R2: 0.999). Furthermore, experiments conducted with the EBRS revealed that the affected limb was capable of accurately tracking the continuous movement of the healthy limb. This synchronization and coordination facilitated efficient upper limb rehabilitation.