Siam-GCAN: A Siamese Graph Convolutional Attention Network for EEG Emotion Recognition
Hong Zeng, Qi Wu, Yanping Jin, Haohao Zheng, Mingming Li, Yue Zhao, Hua Hu, Wanzeng Kong
Abstract
The graph convolutional network shows effective performance in electroencephalogram emotion recognition owing to the ability to capture the brain connectivity. However, the depth information cannot be extracted only through the graph convolutional network structure, and the learning process of the graph convolutional network model ignores the intra-class and the inter-class information. Regarding the above problems, we propose a siamese graph convolutional attention network, named Siam-GCAN, which mainly considers the following two aspects: On the one hand, we use a deep attention layer implemented by multi-head attention mechanism to abstract deeper and valuable features rather than stacking graph convolution layers. On the other hand, we employ the siamese network to cluster the outputs of graph convolutional networks based on Euclidean distance to ensure the learned information has a certain class separability. Experimental results on two public emotional datasets, the SJTU emotion EEG dataset and the SJTU emotion EEG dataset-IV, demonstrate Siam-GCAN outperforms the state-of-the-art baselines in electroencephalogram emotion recognition.