User-Specific Mirror Training Can Improve Myoelectric Prosthesis Control
AbstractState-of-the-art transradial prostheses can provide intuitive and proportional myoelectric control by training an algorithm to correlate surface electromyographic signals from the residual forearm muscles to intended movements of the amputated hand. One training paradigm, “mimicked training,” relies on amputees mimicking a prosthetic hand with their missing hand such that the corresponding muscle activations are correlated to the preprogrammed kinematics of the prosthetic hand. A second training paradigm, “mirrored training,” relies on unilateral amputees mirroring their contralateral hand with their missing hand such that the muscle activations are correlated to the kinematics of the contralateral hand (determined via a motion capture). Prior work with intact participants demonstrated that the kinematics of a given hand are more closely related to that of an individual's contralateral hand as opposed to the preprogrammed kinematics of a prosthesis. This abstract continues our investigation into the training data for myoelectric prostheses by exploring the impact of these training paradigms on real-time prosthetic control with amputees completing a functional task. For one out of three participants, mirrored training significantly improved task performance. These preliminary results demonstrate that mirrored training may provide more dexterous control through task-specific user-chosen training data. These results can guide myoelectric training for proportional and dexterous control.
How to Cite
T. Tully, C. Thomson, G. Clark, and J. George, “User-Specific Mirror Training Can Improve Myoelectric Prosthesis Control”, MEC Symposium, Aug. 2022.
Myoelectric Control Algorithms