Convolutional Neural Network for Heartbeat Classification

Abstract

In recent years, the occurrence of cardiovascular diseases (CVD) has tended to be younger, and the monitoring of abnormal ECG signal is an important ways of preventing CVD. In view of the fact that arrhythmias will only appear in the daily life of patients with a small probability, an ECG signal classification method that fits the actual scene is proposed, which further improves the classification ability of abnormal ECG. Tested by the MIT-BIH arrhythmia database, the overall accuracy of the method reached 92.6%, and the f1 value was 65.9. Compared with the existing methods, the proposed ECG signal classifier is competitive.