Currently, the surface electromyography has gained prominence in human-machine interface systems for prosthetic hand control. However, the absence of tactile feedback often affects their acceptability. To address this problem, this study proposes a closed-loop human-machine interaction system that employs transcutaneous electrical nerve stimulation to reconstruct the hand-brain sensorimotor pathway. Dual-channel surface electromyography signals were synchronously captured from antagonistic muscle groups (flexors/extensors) and processed to generate control commands for the prosthetic hand. Thin film force sensors were applied to record the force signals and encode object stiffness into differentiated transcutaneous electrical nerve stimulation parameters. A stimulation electrode grid was placed on the medial upper arm, targeted the median, ulnar, and radial nerve bundles to evoke tactile sensation on palmar side of the same hand. Participants were instructed to perform object stiffness discrimination across three conditions (soft, medium, and hard) by operating the prosthetic hand with tactile feedback. Analysis of 64-channel electroencephalography signals revealed significant differences in temporal dynamics, spatial distribution, and spectral power of neural responses during object stiffness discrimination. To identify the most accurate classifier, five classifiers were adopted to evaluate the processed electroencephalography data, among which random forest achieved the optimal three-class classification accuracy. This study establishes a highly robust neuroplasticity-inducing control framework for rehabilitation robotics and neuroprosthetics.
Pan et al. (Fri,) studied this question.