Estimation of Ankle Joint Torque and Angle Based on S-EMG Signal for Assistive Rehabilitation Robots

Surface Electromyography (S-EMG) has shown the advantages of robotic rehabilitation. Robotic rehabilitation can be significantly improved if the intended body movement of the patients can be well identified. In this chapter, we first use the SVM classifier to identify the intended motion patterns, which are plantarflexion and dorsiflexion, by using three wireless EMG sensors placed at the tibialis anterior, gastrocnemius lateralis and gastrocnemius medialis muscles. To estimate the ankle joint torque as well as the joint angle for both plantarflexion and dorsiflexion, this chapter also develops nonlinear mathematical models for joint torque estimation and utilises Swarm Techniques to identify model parameters for each movement pattern of the ankle. During rehabilitation, once the intended motion is recognised, the activation functions extracted from an individual associated EMG channel can be used to estimate both the torque and angle by using the established nonlinear models. Experimental results demonstrated the effectiveness of the proposed approach.