Walking speed intention model using soleus electromyogram signal of nondisabled and post-stroke hemiparetic patients

It is well known that the activation of plantar flexors have a strong influence on the walking speed. If the gait speed can be predicted using this relationship, a post-stroke hemiparetic patient could control a gait rehabilitation robot according to his or her gait intention, and the robotic gait rehabilitation effect could be further improved. To find out this relationship, 9 nondisabled subjects and 4 chronic post-stroke hemiparetic subjects performed overground level walking at a comfortable pace, a slow pace, a fast pace, and an increasing pace with electromyogram sensors attached on plantar flexors. Soleus among plantar flexors showed the most stable relationship with walking speed. The relationship between maximum activation level of soleus electromyogram during stance phase before toe-off and walking speed during swing phase after the same toe-off was modeled by a polynomial regression model. The model outputs were then compared to the measured walking speeds using coefficients of determination (R²). The average R² values are 0.594 and 0.692 for 1^st· and 2^nd order models respectively in the nondisabled subjects. The average R² values are 0.598 and 0.623 for the unaffected side and 0.388 and 0.394 for the affected side in the chronic subjects. The results show the feasibility of applying the soleus-walking speed relationship to control the robot gait speed at will. A walking speed estimation method is proposed using only a walking step in real time.