System identification and damping coefficient estimation from EMG based on ANFIS to optimize human exoskeleton interaction

Abstract

although Lower Limb Robotic Rehabilitation device exhibit a great prospect in the rehabilitation of impaired limb, yet it has not been widely applied to clinical rehabilitation due to lack of identification of the system that can build a relationship relating knee joint dynamics to EMG signal. As a result, the insufficient bidirectional information interaction between exoskeleton and patient, the adaptive collaboration is very much absent. In the shared control situation at the interaction point, it is very important that the deficiency of impaired lower limb at knee joint dynamics (Capturing of the intended action of the patient) is extracted beforehand to estimate as to how much assistance need to be provided by the robotic exoskeleton. The intended action data that can be extracted from EMG signal may include the intended posture, intended torque, intended knee joint angle, intended knee joint torque and impedance parameter. In this paper, an application of Adaptive Network based Fuzzy Inference System (ANFIS) has been proposed to identify a proprioceptive feedback system which plays the role of inverse dynamics in the closed loop controlled Robotic Rehabilitation Device. The identified ANFIS inverse dynamic model updates the current status of the interaction force at the patient robotic exoskeleton interaction point. Interaction forces, rate of change in surface electromyography (EMG) signal, Extracted RMS (Root Mean Square) are three input patterns to ANFIS model and impedance parameters damping coefficients, stiffness are the output. The proposed model is able to estimate damping and demonstrate decent accuracy in modulating the knee joint dynamics to minimize the interaction force at the Patient Exoskeleton interaction point.