Muscle-Specific High-Density Electromyography Arrays for Hand Gesture Classification.

Area of Science:

• Biomedical Engineering
• Neuroscience
• Rehabilitation Technology

Background:

• Dexterous hand motion is crucial for object manipulation, and understanding its neural basis requires precise electrophysiological measurements.
• Existing high-density electromyography (HD-EMG) electrode platforms lack muscle specificity, hindering the detailed assessment of intrinsic hand muscles.

Purpose of the Study:

• To develop and evaluate muscle-specific flexible HD-EMG electrode arrays for capturing intrinsic hand muscle activity.
• To assess the utility of these arrays in classifying intended hand motions using machine learning.

Main Methods:

• Developed 60-electrode arrays targeting 10 intrinsic hand muscles for flexible HD-EMG signal acquisition.
• Utilized spatio-temporal amplitude maps to visualize muscle activation patterns.
• Extracted time-domain and temporal-spatial HD-EMG features to train cubic support vector machine classifiers.

Main Results:

• Spatio-temporal EMG maps revealed distinct muscle activation patterns corresponding to specific hand movements.
• Achieved successful classification of 10 common hand motions with average accuracy up to 92±2% using 41 channels.
• Demonstrated that temporal-spatial features enhance motion-specific accuracy, especially for similar muscle recruitment patterns.

Conclusions:

• Muscle-specific electrode arrays effectively record HD-EMG signals from intrinsic hand muscles.
• These arrays enable accurate prediction of hand motions, offering significant potential for research and clinical applications.