Heartbeat ECG Recognition Method for Arrhythmia Classification Via Machine Learning Algorithm

This study presents a comprehensive approach to automatic arrhythmia classification using electrocardiogram (ECG) signals through advanced machine learning techniques. We implemented and compared three deep learning architectures: Convolutional Neural Networks (CNN), Artificial Neural Networks (ANN), and our proposed Weight-based Convolutional Neural Network (WCNN). The WCNN model incorporates wavelet transform for feature extraction, enabling better capture of time-frequency characteristics inherent in ECG signals. Experiments were conducted using a standardized ECG dataset, with signals preprocessed to remove noise artifacts and enhance key features. Performance evaluation metrics included accuracy, sensitivity and specificity across multiple arrhythmia classes. Results demonstrate that the WCNN architecture significantly outperformed both traditional CNN and ANN approaches, achieving higher classification accuracy while maintaining computational efficiency. The WCNN model exhibited particularly strong performance in distinguishing between similar arrhythmia types that pose challenges for conventional algorithms