EEG-based Emotion Recognition Using Spectral Graph Convolutional Neural Network based on Partial Directed Coherence

Due to the recent advent of human-machine interaction, emotion recognition using electroencephalogram (EEG) has gathered significant attention in affective computing. Considering the characteristic of EEG signals, various graph convolutional neural networks have been applied in emotion recognition recently. This work proposes a spectral graph convolutional neural network based on partial directed coherence (SGCN-PDC), which uses the functional connectivity between EEG feature pairs. The result using the SJTU emotion EEG dataset (SEED) showed that the proposed SGCN-PDC performs better than the conventional methods in terms of emotion recognition.