The design of a hemiplegic upper limb rehabilitation training system based on surface EMG signals

Abstract

In this paper, we designed a hemiplegic upper limb rehabilitation training system, which allowed both single degree of freedom and composite degrees of freedom for the training of shoulder and elbow. The system contained an upper limb rehabilitation robot, a pattern recognition system and a motion control system. Firstly, we designed a novel upper limb rehabilitation robot with three degrees of freedom, with the motor and reducer innovatively placed centrally in the shoulder of the mechanical limb arm. The movement is more stable because the inertia of each joint movement is reduced. The design makes simultaneously training both the left and right arm possible. In the control system design, the movement coupling problem is solved through the inverse operation of the target action. Lastly, to further enrich information of the single feature vector, a method integrating the AR model coefficients and wavelet coefficients was proposed. A method combining the Particle Swarm Optimization Algorithm and Levenberg-Marquardt (LM) was used to optimize the BP networks, addressing the problems associated with lower convergence speed and local minimum of standard BP networks. The experiments showed that the convergence speed of the network and the recognition rate of the target action were effectively improved, which demonstrated the effectiveness of the training system.