A new multi-stage combined kernel filtering approach for ECG noise removal

Area of Science:

• Biomedical Engineering
• Signal Processing

Background:

• Electrocardiogram (ECG) signals are often degraded by artifacts and noise, complicating accurate heart disease diagnosis.
• Existing methods struggle to remove noise while preserving essential low-frequency components and subtle ECG features.

Purpose of the Study:

• To propose a novel multi-stage adaptive filtering design for robust ECG signal denoising.
• To preserve low-frequency components and delicate ECG features essential for diagnosis.

Main Methods:

• Development of a combined adaptive filtering approach integrating Kernel Recursive Least Squares Tracker (KRLST) and Approximate Linear Dependency Kernel Recursive Least Squares (ALDKRLS) algorithms.
• Validation using ECG signals from the MIT-BIH database.
• Comparison with existing adaptive filtering techniques.

Main Results:

• The combined ALDKRLS-KRLST approach demonstrated superior performance in attenuating artifacts compared to other methods.
• Enhanced sensitivity in ECG peak detection was observed.
• Improved accuracy in heart disease diagnosis was achieved.

Conclusions:

• The proposed ALDKRLS-KRLST technique offers an effective framework for high-resolution ECG signal processing from noisy recordings.
• This method significantly improves artifact removal and diagnostic accuracy in ECG analysis.