Abstract:
The basis of arrhythmia diagnosis is to correctly identify normal and abnormal individual heartbeat, and classify them into different categories according to ECG waveform. Many existing works rely heavily on data preprocessing, which requires lots of computation and time, and lacks portability between different signals. At the same time, some researches need information input of multiple leads.The residual network (ResNet) is proposed to automatically identify heart rate disorders. In this study, only one lead information is used, and the 4 s heartbeat segment is used as the input data of the classifier to improve the performance of the model by fusing different models. The results show that the accuracy, positive rate and sensitivity of the classifier are 0.855, 0.606 and 0.639, respectively. It is verified that the network based on the model fusion can achieve better arrhythmia classification ability without any artificial preprocessing of ECG signal.