A FEASIBLE ARRHYTHMIA CLASSIFICATION ALGORITHM BASED ON TRANSFORMER MODEL

The auto-classification of arrhythmias plays an essential role in the earlier prevention and diagnosis of cardiovascular disease. Existing deep learning-based methods for arrhythmia classification commonly employ convolutional structures to process spatial information, and employ multiple approaches to process temporal information across data. We propose Arrhythmia Classification Algorithm Based on Transformer Model (CTA) that combines the Transformer model and the Attention mechanism for the prediction of ECG in order to leverage the spatial features and temporal information of the ECG signal. The results of experiments conducted on large datasets in the domain of ECG signals show that the proposed algorithm provides excellent prediction and classification per-formance and serves as a diagnostic aid for doctors.