An Heterogeneous Ensemble Learning based Method for ECG Classification

In this paper, a single beat ECG classification method based on heterogeneous ensemble learning is proposed in order to detect abnormal beats from the Holter ECG. The method extracts features of ECG signals through long short-term memory (LSTM) and 1-D convolutional neural network (CNN). Then we calculate the interval features of ECG signals and fuse two different classification methods by using BP neural network. ECG records from the MIT-BIH arrhythmia database are selected as test data. It is observed that the classification accuracy of the fused model is better than that of the single model. Compared with the classical classification algorithm, our model is not only more efficient than the state of the art in accuracy, but also competitive in terms of sensitivity and specificity.