A Method for Arm Motions Classification and A Lower-limb Exoskeleton Control Based on sEMG signals

Abstract

Exoskeleton robot has been proved to be effective during paraplegia patients rehabilitation. While the exoskeleton is performing a walking gait, patients upper limbs need to exert force to control a pair of crutches in order to keep balance. Danger may occur if the walking gait is not matching movements of the upper limbs. Meanwhile, the exoskeleton should be controlled according to users intention of movement rather than operating deliberately. In this paper, the surface electromyography (sEMG) signals of the patients arms are utilized to offer transparency control interface which comply with human factors. A Back Propagation (BP) neural network of motion recognition method has also been implemented to discriminate seven classes of arm motions. Considering the accurate control commands needed for the exoskeleton robot, we also designed a filter based on the frequency statistics of the number of commands. The results of online experiments exhibit the effectiveness of the proposed approach.