Tuning of robotic therapy controllers for stroke gait: Using isometrically constrained EMG modular structures

Abstract

The development of EMG based controllers for automated gait orthosis is a promising intervention for post-stroke gait rehabilitation and assistive device applications. However, the integration of stroke specific modular neuromuscular patterns for the design of active gait assistance controller is largely unexplored. To this end, this work presents preliminary evidence of the associated variability between lower limb muscle activation patterns underlying discrete isometric control objectives and the analogous dynamic subtask during treadmill gait using matrix factorization algorithms. We find that common muscle coordination patterns underlying both isometric and dynamic gait submovements can be used to reconstruct relevant spatio-temporal patterns for healthy locomotion. We propose that synergies extracted isometrically may be sufficient to tune adaptive controller assistance for disruptive linkages in the EMG control signal to impact gait dysfunction.