A sEMG-based Muscle-Compensation-Error Elimination Module for the Home-Based Bilateral Rehabilitation System

Home-based bilateral training using exoskeletons is considered an effective rehabilitation method to restore the normal motion function of stroke patients' limbs. However, due to the imbalance of limb muscle strength, muscle compensation of the functional upper limb will cause surface electromyography (sEMG) signal fluctuations. The signal fluctuations will lead to prediction errors, thereby affecting the effect of bilateral rehabilitation. To eliminate the prediction error caused by muscle compensation, a novel rehabilitation module for a home-based bilateral rehabilitation robotic system is proposed in this paper. This sEMG-driven rehabilitation module, which incorporates the channel splitting and transferable frequency domain, first identifies the type of muscle compensation mixed in continuous movement and then eliminates the error caused by muscle compensation. Online experiments were conducted to evaluate the system in tracking accuracy and muscle-compensation-error elimination, which showed the proposed rehabilitation system with rehabilitation module achieved fast adaption to the patient's compensatory movement as well as improving the effect of robot-assisted bilateral training.