Electrocardiogram (ECG) signal classification plays a crucial role in medical diagnostics. However, existing Spiking Neural Network (SNN) often exhibit relatively low accuracy when applied to ECG signal classification tasks, limiting their practical applications. To address this issue, this paper proposes a multi-stage model that integrates Convolutional Neural Network (CNN), SNN, and Residual Network (ResNet) to enhance the performance of ECG signal classification. The model first employs CNN to extract spatial features from the ECG signals, followed by SNN to capture temporal sequence information, and finally uses ResNet’s residual block mechanism to improve the learning of deep features. By combining CNN and SNN architectures, the model enhances classification accuracy, while the residual connections in ResNet help mitigate the vanishing gradient problem encountered in deep network training. Additionally, the incorporation of Dropout layers further improves the model’s generalization capability. Experimental results demonstrate that the proposed CNN_SNN_ResNet model significantly improves the accuracy of ECG signal classification, providing a novel approach for the application of Spiking Neural Networks in medical signal processing.