ECG-Based Arrhythmia Classification Using Discrete Wavelet Transform and Attention-Enhanced CNN-BiGRU Model

Electrocardiogram (ECG)-based arrhythmia classification is vital for the early detection and diagnosis of heart diseases. However, raw ECG signals are often noisy, which poses challenges for accurate classification. In this paper, we propose a novel approach that integrates discrete wavelet transform (DWT) for signal denoising with an Attention-Enhanced Convolutional Neural Network-Bidirectional Gated Recurrent Unit (CNN-BiGRU) model for arrhythmia classification. Initially, DWT is applied to the ECG signal to remove noise while preserving important features. To address class imbalance, Borderline-SMOTE is employed to generate synthetic samples for the minority classes. The denoised and balanced signals are then processed through a CNN for feature extraction, followed by a BiGRU to capture temporal dependencies. By incorporating an attention mechanism, the model is able to concentrate on the most critical regions of the ECG signal. The proposed method is evaluated on the MIT-BIH arrhythmia database, achieving an accuracy of 99.22% in classifying five types of arrhythmias, outperforming previous models. This approach provides a promising solution for automatic arrhythmia detection in clinical practice.