Electrocardiogram signals de-noising using lifting-based discrete wavelet transform

Area of Science:

• Biomedical Signal Processing
• Digital Signal Processing
• Wavelet Theory

Background:

• Electrocardiogram (ECG) signals are crucial for diagnosing cardiac conditions.
• ECG signals are often corrupted by nonstationary noises, hindering accurate analysis.
• Existing denoising methods may not effectively handle complex, time-varying noise artifacts.

Purpose of the Study:

• To introduce an effective denoising technique for ECG signals corrupted by nonstationary noises.
• To evaluate the performance of a novel method based on second-generation wavelet transform and level-dependent threshold estimation.
• To compare the proposed method against traditional nonlinear filtering techniques like the median filter.

Main Methods:

• Utilized a second-generation wavelet transform constructed via a lifting scheme for ECG signal decomposition.
• Applied a level-dependent threshold estimator to remove noise from wavelet coefficients.
• Investigated various wavelet filters (Haar, DB4, DB6, Filter(9-7), Cubic B-splines) and decomposition depths.
• Introduced muscle artifact, electrode motion artifact, and white noise for performance evaluation.

Main Results:

• The proposed lifting-based wavelet transform method demonstrated superior denoising performance compared to the median filter.
• Signal-to-noise ratio (SNR) and visual inspection confirmed the effectiveness of the technique across different noise types.
• Performance was evaluated across multiple wavelet filter types and decomposition levels.

Conclusions:

• The developed second-generation wavelet transform technique offers an effective and faster alternative for ECG signal denoising.
• The method shows consistent superiority over median filtering in removing nonstationary noise artifacts.
• This approach enhances ECG signal quality, aiding in more reliable cardiac diagnostics.