Zainul Abidin, Mohammad Yusuf Hamadani, M. Fauzan Edy Purnomo, Syahrul Chilmi
Cardiovascular disease (CVD) is a leading cause of death worldwide, and accurate electrocardiogram (ECG) analysis is essential for early detection. However, ECG signals are often corrupted by noise, including power line interference, and electromyographic artifacts, which compromise accuracy. Wavelet-based denoising offers an effective solution due to its ability to process non-stationary signals, but its performance strongly depends on the choice of mother wavelet and decomposition level. This study compares the denoising performance of seven mother wavelets-sym8, sym5, db4, db6, bior3.7, bior6.8, and coif5-on ECG signals under varying noise conditions. The evaluation employs multiple metrics, including Signal-to-Noise Ratio (SNR), Cross-Correlation (CC), Percent Root Difference (PRD), Mean Squared Error (MSE), and QRS Preservation Score (QPS). Results indicate that coif5 achieves the best noise suppression across most metrics (SNR 4.71 dB, CC 97.11 %) but at the expense of QRS morphology preservation (QPS 93.5%). In contrast, bior3.7 performs poorly at low SNR but excels in preserving QRS features at higher SNR, while bior6.8 shows balanced performance across all metrics (SNR 4.52 dB, CC 96.97%, and QPS 94.30%). These findings emphasize the trade-off between noise reduction and preservation of diagnostic features in ECG denoising and provide a basis for future work evaluating its impact on convolutional neural networks based ECG classification. © 2025 IEEE.
Universitas Brawijaya, Department of Electrical Engineering, Malang, Indonesia; Madiun State Polytechnic, Dept. of Automation Engineering Technology, Madiun, Indonesia; Universitas Brawijaya, Department. of Clinical Pathology, Malang, Indonesia
