EMG Signal Filtering Based on Variational Mode Decomposition and Sub-Band Thresholding

Area of Science:

• Biomedical Engineering
• Signal Processing

Background:

• Surface electromyography (EMG) signals are crucial for analyzing muscle activity but are often corrupted by noise.
• Existing filters typically address only one type of noise, limiting their effectiveness.
• Common noise sources include powerline interference (PLI), baseline wandering (BW), and white Gaussian noise (WGN).

Purpose of the Study:

• To develop and validate a novel filter capable of removing multiple noise types from EMG signals.
• To demonstrate the efficacy of variational mode decomposition (VMD) for comprehensive EMG denoising.
• To enhance the accuracy and robustness of EMG signal processing for downstream applications.

Main Methods:

• EMG signals were decomposed into band-limited modes using variational mode decomposition (VMD).
• Specific noise components (PLI, BW, WGN) were identified and removed within designated modes.
• White Gaussian noise (WGN) was suppressed using soft thresholding with adaptive thresholds.
• Performance was evaluated using simulated and experimental data with metrics like RMSE, SNR improvement, and correlation coefficient reduction.

Main Results:

• The VMD-based filter effectively removed BW and WGN, outperforming traditional methods.
• Significant reduction in PLI noise was achieved, especially at low signal-to-noise ratios (SNRs).
• Average SNR improvements of 18.6 dB (PLI), 19.2 dB (BW), and 8.0 dB (WGN) were observed at -6 dB SNR.
• Experimental results showed complete noise removal in resting states and clear spike detection in action states.

Conclusions:

• Variational mode decomposition (VMD) offers a robust framework for multi-noise removal in EMG signals.
• The proposed VMD-based filter is efficient and suitable for preprocessing EMG data in applications like gesture recognition and decomposition.
• This method provides a significant advancement in EMG signal quality and reliability.