Deep Learning Methods for EEG Applications: Focusing on CNN and RNN

Abstract

In the field of neuroscience, the electroencephalogram (EEG) study aims to discover patterns of various human brain activities in an efficient and accurate manner, which plays an important role in treating brain diseases and other fields. The emergence and development of deep learning networks bring end-to-end approaches towards processing and classifying human brain signals. In the first section of the study, two typical networks used in EEG applications—the convolutional neural network (CNN) and the recurrent neural network (RNN)—are introduced, then the study measures the performance of each network in relation to their specific characteristics in temporal and spatial domain respectively and analyzes how noise and other interference towards EEG signals affect the training of the two networks. To make an integral view of how to take advantage of both networks mentioned above for EEG applications, the paper finally introduces a general system framework using a combination of CNN and RNN. This study will bring important value to the research and application of deep learning methods for EEG applications.