Thresholding noise reduction algorithm for ECG signals based on wavelet analysis

Electrocardiogram (ECG) signals are essential for diagnosing heart diseases but are often degraded by high-frequency noise, baseline drift, and electromyographic interference. Traditional denoising methods often fail to remove noise while preserving key signal features. Wavelet transform combined with thresholding provides an effective alternative, enabling signal decomposition in the time-frequency domain to isolate and suppress noise. This study systematically compares fixed threshold methods (soft and hard threshold) and adaptive strategies (VisuShrink, SureShrink, BayesShrink) across three wavelet bases (db4, sym5, coif3), alongside traditional filters (moving average, adaptive filtering). Results show that Soft Threshold excels in high-frequency noise removal, BayesShrink is most effective against baseline drift, and SureShrink performs best for electromyographic interference. Hard thresholding showed weaker performance. Among wavelet bases, coif3 preserved signal details slightly better. Combining it with adaptive thresholds is recommended for practical ECG denoising. This approach enhances ECG signal quality and supports reliable clinical and remote health monitoring.