Convolutional Neural Networks with A Topographic Representation Module for EEG-Based Brain-Computer Interfaces

Objective: Convolutional Neural Networks (CNNs) have shown great potential in the field of Brain-Computer Interface (BCI) due to their ability to directly process the raw Electroencephalogram (EEG) without artificial feature extraction. The raw EEG signal is usually represented as 2-Dimensional (2-D) matrix composed of channels and time points, which ignores the spatial topological information of EEG. Our goal is to make the CNN with the raw EEG signal as input have the ability to learn the EEG spatial topological features, and improve its classification performance while essentially maintaining its original structure. Methods: We propose an EEG Topographic Representation Module (TRM). This module consists of (1) a mapping block from the raw EEG signal to a 3-D topographic map and (2) a convolution block from the topographic map to an output of the same size as the input. We embed the TRM to 3 widely used CNNs, and tested them on 2 different types of publicly available datasets. Results: The results show that the classification accuracies of the 3 CNNs are improved on both datasets after using TRM. The average classification accuracies of DeepConvNet, EEGNet and ShallowConvNet with TRM are improved by 4.70\%, 1.29\% and 0.91\% on Emergency Braking During Simulated Driving Dataset (EBDSDD), and 2.83\%, 2.17\% and 2.00\% on High Gamma Dataset (HGD), respectively. Significance: By using TRM to mine the spatial topological features of EEG, we improve the classification performance of 3 CNNs on 2 datasets. In addition,since the output of TRM has the same size as the input, any CNN with the raw EEG signal as input can use this module without changing the original structure.