Human emotion recognition based on multi-channel EEG signals

EEG is a kind of sequential data. In order to learn the relationship of EEG in time series, this paper proposes to use long-term memory network to combine convolutional neural network and self-attention mechanism to extract features from EEG. In order to verify the superiority of this method, I design a complete experiment based on SEED data set, and the experimental strategy is an independent leave one method cross-validation experiment for subjects. The experimental results show that the proposed method can effectively learn the EEG temporal relationship, and can improve the convergence speed and have a high accuracy.