A study on the universal method of EEG and ECG prediction

As one of the most important attribute of nonlinear dynamic system, chaotic time series include electroencephalogram (EEG) and electrocardiogram (ECG) have been widely studied for decades. However, the universal prediction method of them is still unexplored due to their inherent random feature and complexity. Considering the high-layer information of images and time-correlation of time series data, traditional support vector machine (SVM), convolution neural network (CNN) and bi-directional recurrent neural network (BRNN) are the main models being used. In this work, by combining CNN with BRNN, we developed a universal model (BRCNN) for high accurate prediction of two of chaotic time series problems (CTSPs), EEG and ECG. For comparison, three models include SVM, CNN, and BRCNN are simultaneously performed on a dataset of EEG signal and a dataset of ECG signal, results demonstrated a superior classification quality of BRCNN (i.e., ~0.90 and ~0.85 AUC, respectively). Such a high prediction accuracy of BRCNN provides the possibility of applying a universal model for high accurate prediction of EEG and ECG.