EMG신호기반 특징추출과 근육부위 선택을 통한 계단보행 단계 분류

초록

In this paper, we propose a EMG signal based four-step stair gait phase recognition method using feature and muscle selection. A lot of existing stair walking studies were focused on the center of gravity. AHRS sensor is used to train the LDA classifier for four-step ascent/descent gait phase. Experimental results show that the average recognition rate of ascent is about 64% when Vastus Medialis and Semitendinosus are selected and VAR, WAMP features are used. In case of descent, the recognition rate is lower than that ascent but it shows similar patterns according to muscle and feature.