Leveraging Motor Unit Spatial Activation Patterns for Channel Selection in Finger Force Regression.

Area of Science:

• Biomedical Engineering
• Neuroscience
• Rehabilitation Engineering

Background:

• High-density surface electromyography (sEMG) signals enable prediction of individual finger forces.
• This capability is crucial for advanced human-machine interfaces, particularly in prosthetic limb control.
• However, the high electrode count in sEMG systems presents significant computational challenges.

Purpose of the Study:

• To develop and evaluate a-priori channel selection strategies for sEMG-based finger force decoding.
• To reduce computational requirements by minimizing the number of sEMG channels used.
• To test the generalizability of pre-selected channels across different subjects, contrasting with subject-specific methods.

Main Methods:

• Utilized a-priori channel selection guided by motor unit spatial activation patterns.
• Compared the performance of a reduced subset of 32 sEMG channels against a full 256-channel setup.
• Evaluated decoding accuracy using Root Mean Square Error (RMSE) as a percentage of Maximum Voluntary Contraction (MVC).

Main Results:

• A subset of 32 sEMG channels achieved an RMSE of 6.32 ± 2.34% MVC.
• This performance is competitive with the state-of-the-art baseline using all 256 channels (5.57 ± 1.94% MVC).
• The a-priori selection strategy demonstrated effectiveness without compromising decoding accuracy.

Conclusions:

• Simple, a-priori sEMG channel selection is a viable strategy for decoding finger forces.
• This method significantly reduces computational demands, making it suitable for resource-limited applications.
• The findings support the use of generalized channel selection for prosthetic control and other human-machine interfaces.