Evaluation of Hand Gesture Classification Based on Motor Units Activation for Prosthetic Application

Area of Science:

• Biomedical Engineering
• Neuroscience
• Rehabilitation Technology

Background:

• Advanced upper-limb prosthetics utilize electromyography (EMG) pattern recognition for control.
• Current EMG feature extraction methods often overlook the biological origin of signals.
• Motor unit decomposition offers a biologically relevant alternative for prosthetic control.

Purpose of the Study:

• To evaluate the feasibility of using motor unit-based features for classifying hand gestures.
• To compare the accuracy of motor unit features against traditional root-mean-square (RMS) features in prosthetic control.
• To identify limitations for real-world application of motor unit-based prosthetic control.

Main Methods:

• Recorded high-density EMG signals from 5 participants performing 14 hand gestures.
• Decomposed EMG signals using blind source separation to identify motor unit activity.
• Computed motor unit features and applied a classification paradigm, comparing it to RMS-based classification.

Main Results:

• Motor unit-based classification achieved accuracies between 80.6% and 90.3%.
• In some cases, motor unit features yielded accuracy comparable to or exceeding RMS-based methods.
• Motor unit classification showed higher inter-subject variability than RMS-based methods.

Conclusions:

• Motor unit features demonstrate potential for decoding hand gestures in prosthetic control.
• Technological and methodological advancements are necessary for practical implementation.
• Further research is crucial to translate these findings into effective prosthetic applications.