Identification and Classification of Electrocardiogram Signals Based on Convolutional Recurrent Neural Network

Abstract

Electrocardiogram (ECG)signals are important sign signal of human heart health. Arrhythmia is one of the main features of heart disease. Therefore, ECG signal recognition and classification have important clinical significance. In this paper, the ECG signals in the MIT - BIH standard library were used as sample data, which were identified and classified based on the algorithm of convolutional recurrent neural network (CRNN)in order to realize the intelligent identification and classification of ECG signals. The R-wave peak location and heartbeat segmentation of the ECG signals were performed on the sample data using the differential threshold method, and a convolutional recurrent neural network was constructed to identify and classify the signals. The classification results show that the overall recognition rate of ECG signals in the MIT - BIH database sample is 98.81 %, the recognition rate of normal ECG signals is up to 99.67%. The results show that the CRNN has strong generalization ability, fast convergence rate and a good recognition classification rate for ECG signals.