A hybrid complex-valued neural network framework with applications to electroencephalogram (EEG)

Abstract

In this article, we present a new EEG signal classification framework by integrating the complex-valued and real-valued Convolutional Neural Network (CNN) with discrete Fourier transform (DFT). The proposed neural network architecture consists of only one complex-valued convolutional layer, real-valued convolutional layers, and fully connected layers. Our method can efficiently utilize the phase information contained in the DFT. We validate our approach using two simulated EEG signals and two benchmark datasets and compare it with some widely used frameworks. Our method drastically reduces the number of parameters used and improves accuracy when compared with the existing methods in classifying benchmark seizure EEG dataset, and significantly improves performance in classifying simulated EEG signals.