Arm Motion Classifiction Based on sEMG and Angle Signal for A Lower Limb Exoskeleton Control System

When walking with the lower limb rehabilitation exoskeleton, the wearers should control the exoskeleton to switch gait by their upper limbs recur to buttons on the crutch. Since both hands are occupied for support, this on human operation relied control strategy is extremely inefficient and discomfort. Otherwise, control of the exoskeleton should be the wearer's movement intention rather than their operation based, because operation means time delay, which makes the switching of exoskeleton gait discontinuous. In this paper, a new surface electromyography (sEMG) and angle signals called EA-CNN (EMG and Angel Combined Neural Network) based neural network is proposed to accomplish the classification of wearers' arm movement intentions, this represents the switching signal of exoskeleton gait. Compared with the classification method using only sEMG signal, EA-CNN shows better robustness in motion recognition. The experimental results show that the exoskeleton can be controlled accurately with this control strategy.