Multi-Feature Fusion Graph Convolutional Network Based on Domain Adaptation for EEG Emotion Recognition

Emotion recognition based on EEG signals facilitates the timely detection of changes in the emotional state of personnel. However, due to personal differences, it is not easy to use the model trained on one individual directly for the emotion recognition of other users. Meanwhile, the current primary EEG signal acquisition devices are 62-conductor acquisition caps, which need to be more convenient and efficient in practical use. For this reason, we propose a graph convolutional network with multi-feature fusion and use the domain adaptation module to improve the model’s adaptability. The multi-feature fusion is also used to improve the multi-feature information of the model. Many experiments show that our proposed method improves the accuracy by 25.7% and reduces the electrodes by 97%.