基于双分支特征融合和交叉注意力网络的脑电图情绪识别

打开文本图片集

中图分类号：TP391文献标识码：A

Abstract： To effectively integrate utilize the rich features embedded in EEG signals for enhancing emotion recognition performance，a Dual-Branch Feature Fusion Cross-Attention Network （DBF-CANet） is proposed. The model combined a Transformer with Convolutional Neural Network （CNN） designed a CrossAttention Fusion Module （CAFM） tailored to the complementary differential nature dual-branch features，achieving efficient feature fusion. Considering the temporal dynamics EEG signals，a time attention module based on Long ShortTerm Memory （LSTM） was further introduced to explicitly capture long-range temporal dependencies within the signals. Rigorous evaluation on the DEAP SEED benchmark datasets validated the model's effectiveness. Experimental results demonstrate the model's outsting performance in emotion recognition tasks.

Keywords： EEG signals；emotion recognition； Transformer； feature fusion；cross-attention

情感是人的情绪体验和心理反应，作为人类心理活动的重要组成部分，对个体的认知、行为和生理功能有重要影响[1]。（剩余10625字）