Hierarchical Motion Segmentation through sEMG for Continuous Lower Limb Motions

Abstract

Surface electromyogram (sEMG), an electrical signal generated from muscles, has been used for a long time to decode human motion intentions for interactions between a robot and human. To support not only the diverse movements in human daily living but also the task of increasing the human' robot interface and its applications, a new algorithm that can classify continuous lower limb motions using sEMG signals is proposed herein. By simply constructing motion hierarchy and probability distribution of sEMG for each motion phase obtained by using only kinematic motion data and sEMG data, it is possible to demonstrate higher classification accuracy than state-of-the-art supervised learning methods consuming much time. Four different experiments were performed on five participants and the algorithm was verified to successfully distinguish between walking from running, and between standing up and sitting down from jumping.