Prediction Method of Walking Speed at Swing Phase using Soleus Electromyogram Signal at Previous Stance Phase.

A recent research has proposed a prediction method of walking speed with soleus electromyogram (EMG) signal activation level at push-off phase. However, the prediction of walking speed at low speed is inaccurate and the coefficients of determination (R values) of the used linear regression model is low. In this study, we propose a new method for predicting walking speed during swing phase with soleus EMG signal activation levels at pre-load and push-off phases, and square root value is used as a feature. The proposed method is verified by walking experiment with 5 nondisabled subjects. (R values) of the new method is improved by 10.3 % than that of the method used in the previous study. And the proposed method improves accuracy mainly at low speed and precision at high speed to predict a correct walking speed throughout walking speed range. Thus, the proposed method enhances the performance of the prediction model of walking speed without being biased in the range of high or low speed. The proposed method has potential to be used to control the gait speed of a lower-limb exoskeleton according to wearer's gait intention.