EEG Emotion Recognition Based on Dynamical Graph Attention Network
Yi Guo, Chao Tang, Hao Wu, Badong Chen
Abstract
Emotion recognition based on electroencephalography (EEG) signals is one of the current research challenges in this field. In order to learn the optimal graph structure information for each subject, we propose a dynamic graph attention neural network model. The model utilizes a graph attention neural network as a feature learner, dynamically learning channel connections, and enriching feature representations between channels through global attention. To verify the effectiveness of the proposed method, we conducted experiments on the publicly available emotion recognition dataset SEED. The experimental results show that the average accuracy and standard deviation of the 15 subjects are 94.6% and 4.98%, respectively. The results indicate that our proposed dynamic graphical attention neural network outperforms existing methods.