A Wavelet-Based Architecture for Efficient ECG Signal Denoising

Abstract

Purpose: An electrocardiogram (ECG) is one of the most important biomedical signals in the detection and diagnosis of heart arrhythmias. As an interpretable biomedical signal, the ECG is subject to various interferences and noises, such as: baseline drift, power-line noise, and white Gaussian noise; all of which may obscure vital information and diagnostic features. This study aims to devise an efficient and simple method for noise reduction in ECG signals, to enhance the clarity of the signal while retaining the morphologies of the waveforms. Approach: This study supports the simple technique of single Discrete Wavelet Transform (DWT) based architecture to signal decomposition and threshold adjustments, a straightforward method. This technique was tested on both synthetic ECG signals and ECG recordings from the MIT-BIH database. The performance of this method was analyzed based on the correlation coefficient (CC) and signal-to-noise ratio (SNR) in relation to older signal-cleaning techniques. Results: The implemented system attained excellent performance, as indicated by the CC values of 0.9934, 0.9832 and 0.9524 in relation to the power-line noise, baseline drift, and white Gaussian noise, respectively. The SNR improved by 17.97 dB, 15.45 dB, and 10.06 dB, surpassing the previous methods of 16.58 dB, 14.82 dB, and 7.80 dB.  Conclusion: The outcomes demonstrate that DWT technique is efficient in minimizing multiple noise types in one operation, which ultimately enhances performance of the ECG signal. This improvement in SNR and waveform link supports its use in correct clinical diagnosis and real-time biomedical uses.