A Spatio-Temporal Graph Neural Network for EEG Emotion Recognition Based on Regional and Global Brain

EasyChair Preprint no. 12453

Abstract

Effective emotion recognition based on electroencephalography (EEG) is of relevant importance for the investigation of intelligence of the Brain-Computer Interface (BCI). Neuroscientific studies suggest that investigating localized brain activities contributes to a deeper understanding of the functionality of specific brain regions and the activity patterns under different emotional states. Many deep learning-based methods have been employed for EEG emotion recognition in recent years; however, most of these methods fail to extract the spatio-temporal features of EEG signals adequately. To further improve the efficiency of EEG emotion recognition, we propose in this work a novel spatio-temporal graph neural network, namely MSL-TGNN, by integrating local and global brain information. That is, the multi-scale temporal learner is employed to extract temporal features of EEG data. To explore the spatial features of EEG signals, considering the varying roles of different brain regions in EEG emotion classification, we propose a brain region learning block and an extended global graph attention network. The brain region learning block aggregates local channel information, and the extended global graph attention network can effectively capture nonlinear dependencies among regions and global brain information, thereby enhancing the learning capability for the EEG data. We conducted subject-dependent and subject-independent experiments on the DEAP dataset, and the results obtained indicate that our proposed model outperforms compared to state-of-the-art methods.

Keyphrases: Bidirectional Gated Recurrent Unit, deep learning, EEG emotion recognition, Graph Attention Network

@Booklet{EasyChair:12453,
  author = {Xiaoliang Wang and Chuncao Li and Yuzhen Liu and Kuan-Ching Li and Wei Liang and Shiqi Zheng},
  title = {A Spatio-Temporal Graph Neural Network for EEG Emotion Recognition Based on Regional and Global Brain},
  howpublished = {EasyChair Preprint no. 12453},

  year = {EasyChair, 2024}}