Muscle Synergy-based Grasp Classification for Robotic Hand Prosthetics.

Area of Science:

• Biomedical Engineering
• Robotics
• Neuroscience

Background:

• Robotic prosthetics require intuitive control methods.
• Surface electromyography (EMG) offers a non-invasive way to capture muscle activity.
• Muscle synergies represent coordinated muscle activation patterns.

Purpose of the Study:

• To evaluate the efficacy of muscle synergies derived from EMG for hand posture classification in prosthetic control.
• To compare Non-negative Matrix Factorization (NMF) and Principal Component Analysis (PCA) for extracting muscle synergies.
• To assess the robustness of these methods when electrode data is incomplete.

Main Methods:

• EMG signals were recorded from intact-limb subjects.
• Muscle synergies were extracted using NMF and PCA.
• Hand posture classification was performed using features derived from these synergies.
• Muscle importance was evaluated by simulating electrode removal.

Main Results:

• NMF-derived muscle synergy features significantly outperformed PCA-derived features for hand posture classification.
• NMF demonstrated robustness in classifying hand postures even with simulated electrode loss.
• The study successfully classified 14 distinct hand gestures, including a resting state.

Conclusions:

• Muscle synergies extracted via NMF are a viable and robust method for hand posture classification in EMG-based prosthetic control.
• NMF offers advantages over PCA in terms of feature extraction and robustness for this application.
• This approach holds promise for improving the functionality of upper-limb robotic prosthetics.