Practical method for predicting intended gait speed via soleus surface EMG signals

Abstract

The lack of patient effort during robot‐assisted gait training (RAGT) is thought to be the main factor behind unsatisfactory rehabilitative efficacy among hemiparetic stroke patients. A key milestone to implement patient‐driven RAGT is to predict gait intent prior to actual joint movement. Here, the authors propose a method of predicting step speed intent via surface electromyogram (EMG) signals from the soleus. Six lower‐limb muscles were initially evaluated on a treadmill, and the results suggest that the soleus EMG signals correlate well with step speed. The authors further propose a simple linear regression model which predicts subsequent step speed via current soleus EMG signals with over‐ground gait sessions, of ∼0.6. The proposed experimental results and simple prediction model should be applicable for RAGT without significant modifications.