Mirror-Training of a Cable- Driven Hand Rehabilitation Robot Based on Surface Electromyography (sEMG)

In recent years, robots are widely used for helping post-stroke patients do rehabilitation training because it can provide long-term, accurate stimulation for motor function recovery. However, how to design a useful robot that can help patients do rehabilitation training such as separate movements and how to establish a human-robot interaction interface to increase the patient's involvement are challenging topics for the hand rehabilitation robot. Therefore, a hand exoskeleton robot has been designed to help the post-stroke patient do hand rehabilitation training with the aid of some advanced control methods. There are two notable features on this robot: 1) the active disturbance rejection controller is utilized to control the robot for a better control performance. Experimental results show that this controller can track the reference better than PID controller and can reject the disturbance as well; and 2) this paper creates a human-robot interaction interface to do active rehabilitation control (mirror-training). Firstly, this paper utilizes the back-propagation neural network to recognize the volunteer's movement intentions (hand gestures) based on surface electromyography (sEMG). Then, the corresponding hand-gesture recognition result is used to control the hand exoskeleton. The result shows that the rehabilitation robot can follow the volunteer's movement intention to fulfill the mirror-training of the patient.