Surface Electromyography Characteristics for Motion Intention Recognition and Implementation Issues in Lower-limb Exoskeletons

Abstract

Recognizing the user's motion intentions is a crucial challenge to develop human augmented robotic devices due to safety and easiness of interactions. Among the possible sensorial modalities, surface electromyography (sEMG) signals have been tested to be a primary motion intention channel due to the inherent advantage of electromechanical delay and the muscle activation information. However, the lack of detailed sEMG characteristics as motion recognition has been difficult issues to develop safe and intuitive interactions with the robots. In this study, we evaluated the sEMG characteristics for their potential applicability to recognizing the motion intentions of humans. For the discrete motion intention recognition, the walking environments were classified using only sEMG signals by support vector machine (SVM) and linear discriminated analysis (LDA) models with accuracy of 79.1% and 76.3%. Due to the fact that it is crucial to identify an unexpected disturbance by the collision between the exoskeleton and surrounding environment in recognizing the user intention to guarantee the safety of the user, sEMG and torque sensors were used to classify user-intended interaction forces and disturbance forces in the event of collisions. A control algorithm was proposed that detects and compensates for collisions, and its performance showed that robust motion intention recognition and control of powered exoskeletons are possible. We investigated the effect of muscle fatigue caused by long-term walking with heavy load wearing an exoskeleton. The sEMG amplitude and frequency were analyzed for muscle fatigue due to single-joint (knee extensions) and multi-joint (walking) exercises, and muscle fatigue due to walking was prominent in the signal from the vastus medialis (VM). The characteristics of sEMG due to muscle fatigue should be seriously considered in continuous motion estimation.