ECG Signal Classification Based on Attention Mechanism and 1D-CNN

Atrial fibrillation (AF) is a prevalent clinical arrhythmia, posing a significant health risk. Efficient diagnosis relies on ECG signals. To enhance timely and accurate AF diagnosis, we propose a method that integrates attention mechanism and convolutional neural network (CNN) for ECG signal classification. Firstly, we improve signal-to-noise ratio via discrete wavelet transform. Next, we extract features using 1dCNN. Finally, the attention mechanism assigns different weights to extract important information, enhancing classification performance and generalization. Extensive comparative and ablation experiments validate our method's effectiveness, achieving 99.60% accuracy and 99.61% F1-Score on the test set. This approach offers a superior tool for ECG signal classification and diagnosis.