Control of exoskeleton using EMG sensors

Abstract

Exoskeletons (Exo) are bio-robots that are widely used to enhance wearers’ performance or to assist disabled people in mobility and rehabilitation. One of the major problems faced by Exo developers is how to predict the required joint moving parameters, such as joint torques, joint angular displacements or velocities that must be fed to the motion controller. EMG signals from human muscles offer opportunities to solve such the problem. This paper introduces a sensor-based system that can predict the required joint torque on the basis of two EMG signals and an angular displacement signal. The EMG signals are detected from the two muscle groups: quadriceps and hamstrings. The system is combined from a commercial DAQ and an adaptive neuro-fuzzy inference system (ANFIS) developed by the authors. The outputs of the ANFIS are then used to control the knee motion. Experimental results showed that performance of the people wearing this EMG-based Exo was significantly improved.