An ECG Signal Denoising Method Using Filtering Techniques

Abstract

This study presents a novel approach for denoising electrocardiogram (ECG) signals, aimed at improving their performance and availability under noisy conditions.The suggested method harnesses Conditional Generative Adversarial Networks (CGANs) tailored for the purpose of denoising Electrocardiogram (ECG) data. This approach comprises two essential elements: an improved Convolutional Auto-Encoder (CAE)-driven generator and a discriminator with four convolutional layers and a fully connected layer. It takes advantage of the ECG signal's innate ability to retain spatial proximity and neighboring patterns in more advanced feature representations, aided by skip connections that aid gradient propagation during the denoising training process, the resulting model exhibits strong performance and generalization capabilities. This method incorporates advanced filtering techniques, including IIR notch filters, The combination of l₂ and l₁ trend filtering along with a Kalman filter has been employed to enhance the quality of an ECG signal, rendering it more amenable for subsequent analysis and identification. These filtering methods prove particularly advantageous when dealing with ECG signals exhibiting a high Signal-to-Noise Ratio (SNR). Extensive experimentation conducted using the MIT-BIH database has validated the efficacy of these filtering techniques in effectively eliminating various sources of noise, while preserving the unique characteristics of the ECG signals.