The Effectiveness of Optimal Discrete Wavelet Transform Parameters Obtained Using the Genetic Algorithm for ECG Signal Denoising

Abstract

The analysis of electrocardiogram (ECG) signals is imperative for the diagnosis of cardiac anomalies. However, the integrity of ECG signals is often compromised by the presence of noise, such as Additive White Gaussian Noise (AWGN) and Power Line Interference (PLI), which can obfuscate critical signal characteristics during acquisition. AWGN typically permeates ECG recordings through electronic noise, movement artifacts, and environmental factors, whereas PLI is commonly induced by alternating current power sources at frequencies of either 50 or 60 Hz, contingent upon the geographical location. In this investigation, a novel denoising strategy employing a synergistic application of the Genetic Algorithm (GA) and Wavelet Transform (WT) is presented. The WT parameters are meticulously optimized through the Genetic Algorithm, which conducts a systematic search to ascertain the optimal decomposition levels and thresholding values for noise reduction. This iterative optimization process refines WT parameter settings to attenuate noise effectively. The efficacy of the proposed approach is rigorously evaluated using the benchmark MIT-BIH Arrhythmia Database, a renowned and publicly accessible collection of annotated ECG recordings. Objective metrics, namely the Signal-to-Noise Ratio (SNR) and the Percentage Root mean square Difference (PRD), are utilized to validate the performance enhancements achieved by the proposed method. Results indicate that the method substantially mitigates both PLI and AWGN, yielding a cleaner ECG signal that is more amenable to subsequent medical analysis. Notably, for PLI at 50 Hz with an input SNR of 10 dB, the algorithm achieved an output SNR of 22.56 dB and a PRD of 7.46%. Similarly, under AWGN conditions with an equivalent input SNR, an output SNR of 18.31 dB and a PRD of 12.16% were realized. These outcomes signify a notable improvement over existing methodologies documented within the literature, affirming the proposed method's potential for advancing ECG signal processing in medical applications.