Automated Channel Selection in High-Density sEMG for Improved Force Estimation.

Area of Science:

• Biomedical Engineering
• Neuroscience
• Rehabilitation Engineering

Background:

• Accurate force estimation from surface electromyogram (EMG) signals is crucial for various applications.
• High-density (HD) EMG offers rich data but poses challenges for real-time processing and dimensionality.

Purpose of the Study:

• To develop and validate novel methods for selecting optimal subsets of HD-EMG channels.
• To improve force estimation accuracy and reduce dimensionality for real-time applications.

Main Methods:

• Collected isometric contraction data (HD-EMG and force) from biceps brachii and brachioradialis.
• Employed Fast Orthogonal Search (FOS) for force estimation.
• Utilized frequency-domain Principal Component Analysis (PCA) and a novel Power-Correlation Ratio (PCR) for channel selection.

Main Results:

• Achieved up to 30% improvement in force estimation accuracy.
• Reduced data dimensionality by up to 57% using selected channel subsets (1-3 channels per muscle).
• Demonstrated superior performance of PCA and PCR over time-domain PCA for channel selection.

Conclusions:

• Novel channel selection techniques (frequency-domain PCA, PCR) effectively reduce HD-EMG dimensionality.
• Optimized channel subsets significantly enhance real-time force estimation accuracy.
• These findings support more efficient and accurate EMG-based force prediction systems.