Error reduction in EMG signal decomposition.

Area of Science:

• Biomedical Engineering
• Neuroscience
• Signal Processing

Background:

• Decomposition of electromyographic (EMG) signals into motor-unit action potentials is crucial for understanding muscle control.
• Automated and manual decomposition methods are prone to errors, including location and identification inaccuracies, especially with ambient noise.
• Existing methods require refinement to enhance the precision of motor-unit firing instance identification.

Purpose of the Study:

• To classify and reduce errors in motor-unit action potential train (MUAPT) decomposition from surface EMG (sEMG) signals.
• To develop and evaluate an algorithm that combines multiple decomposition estimates for improved accuracy.
• To assess the algorithm's performance in reducing identification and location errors in motor-unit firing instances.

Main Methods:

• Analysis of 1,061 MUAPTs from sEMG signals recorded during human voluntary contractions.
• Classification of decomposition errors into location (temporal variability) and identification (missed/false detections) categories.
• Development of an error-reduction algorithm leveraging multiple decomposition estimates to achieve a more probable firing instance identification.

Main Results:

• The developed error-reduction algorithm successfully reduced identification errors by an average of 1.78%, achieving 97.0% accuracy.
• Location errors were reduced by an average of 1.66 ms.
• Algorithm performance involves a trade-off between MUAPT yield and decomposition time.

Conclusions:

• The proposed error-reduction algorithm effectively mitigates decomposition errors in sEMG analysis.
• This method enhances the precision of identifying motor-unit firing instances, crucial for applications like synchronization analysis.
• The algorithm is versatile and can be integrated with various decomposition strategies.