Electromyographic (EMG) control interfaces have the potential to increase the effectiveness and accessibility of rehabilitation robotics to a larger population of impaired individuals, including those with no residual motion in their upper limb. Building on our previous work to characterize the surface EMG patterns of able-bodied and incomplete spinal cord injury (iSCI) subjects, we have developed a real-time controller for the MAHI EXO-II upper limb exoskeleton. This controller uses a pattern recognition approach to detect the user’s intended direction of motion when presented with a voluntary isometric contraction by the user and triggers appropriate motion of the robot in response. The controller is trained to work in four single degree-of-freedom (DoF) modes corresponding to the elbow and wrist joints as well as two multi-DoF modes that are each a combination of two of the single DoFs. Real-time control has been evaluated experimentally with able-bodied and iSCI subjects in a single-session protocol that included calibration and training of the pattern recognition algorithm, followed by two modes of testing – with and without robot motion. The protocol was made automatically adaptive in 1) the EMG activity required to trigger classification of intended direction, 2) the amount of training data that each mode received, and 3) the specific electrodes and specific features of the EMG waveform that were used as inputs to the classification algorithm. The final result is a highly functional and intuitive system capable of training users with differing physiology to generate specific muscle activity in a desirable and repeatable way. In addition to the overall classification performance, the subject-specific aspects of the controller are being compared across both able-bodied and iSCI subjects to better understand its clinical potential.
, “Characterization of Surface Electromyography Patterns of Healthy and Incomplete Spinal Cord Injury Subjects Interacting with an Upper-Extremity Exoskeleton”, International Conference on Rehabilitation Robotics (ICORR). IEEE, London, UK, 2017.