Data Prediction of ECG Based on Phase Space Reconstruction and Neural Network

摘要

A new data prediction method for electrocardiogram (ECG) signals is proposed, which combines phase space reconstruction theory and back propagation neural network (BPNN). The proposed method involves three parts. First, two key parameters for phase space reconstruction are solved: delay time by mutual information method, and embedding dimension by Cao method. Second, phase space of ECG signals is reconstructed based on the two solved parameters, then the reconstruction data is input to BPNN, and the input layer structure of neural network is determined. Finally, the BPNN is trained using the reconstructed data, and the prediction model is established to complete the data prediction of ECG signal. Simulation results indicate that the proposed method performs well in ECG signal prediction and has great reference value for ECG signal data prediction.