An sEMG Denoising Method with Improved Threshold Estimation for Rapid Keystroke Tasks

Area of Science:

• Biomedical Engineering
• Signal Processing
• Neuroscience

Background:

• Surface electromyography (sEMG) signal acquisition is prone to noise, compromising data quality and reliability.
• Existing denoising techniques often struggle with high-duty-cycle sEMG signals, such as those from rapid keystrokes, due to biased noise estimation.
• This limitation hinders accurate analysis of myoelectric activity during repetitive, high-rate movements.

Purpose of the Study:

• To develop an advanced sEMG denoising method for enhanced signal processing during rapid keystroke tasks.
• To address the limitations of conventional denoising methods in handling high-duty-cycle sEMG signals.
• To improve the reliability and accuracy of sEMG analysis in applications involving fast, repetitive motions.

Main Methods:

• Integration of the Walrus Optimizer (WO) with Variational Mode Decomposition (VMD) to optimize VMD parameters (K and α).
• Development of an improved threshold estimation strategy tailored for high-duty-cycle sEMG signals during rapid keystrokes.
• Validation using sEMG data from 18 participants performing rapid keystroke tasks across various signal-to-noise ratios (SNRs).

Main Results:

• The proposed WO-VMD method achieved significant improvements in signal-to-noise ratio (ΔSNR) of 2.75–6.65 dB.
• A substantial reduction in root-mean-square error (ΔRMSE%) ranging from 27% to 53% was observed.
• Spectral fidelity was maintained, with median frequency variation rate (ΔMDF%) below 3.48%.

Conclusions:

• The proposed integrated WO-VMD method offers an efficient and reliable solution for denoising sEMG signals.
• This technique effectively overcomes the challenges posed by high-duty-cycle muscle activation in rapid keystroke analysis.
• The findings support the application of this advanced denoising approach for improved myoelectric signal processing.