文献类型：期刊文献

英文题名：The mitigation of heterogeneity in temporal scale among different cortical regions for EEG emotion recognition

作者：Xu, Zhangyong[1];Chen, Ning[1];Li, Guangqiang[1];Li, Jing[1];Zhu, Hongqing[1];Zhu, Zhiying[1]

机构：[1]East China Univ Sci & Technol, Sch Informat Sci & Engn, Shanghai 200237, Peoples R China

年份：2025

卷号：309

外文期刊名：KNOWLEDGE-BASED SYSTEMS

收录：;EI(收录号：20245017512245);WOS:【SCI-EXPANDED(收录号:WOS:001385579600001)】；

基金：Acknowledgments This work was supported by the National Natural Science Founda-tion of China [grant number 61771196, 61872143] . We would like to thank the authors of Ref. [24,39] for providing the codes of their models and helps for us.

语种：英文

外文关键词：Electroencephalogram (EEG); Emotion recognition; Heterogeneity; Neuroscience knowledge

摘要：Neuroscience studies suggest that, as high-level cognitive processes, various emotional states will lead to cooperative activation among different cortical regions, which process information across diverse temporal scales. However, such heterogeneity existing in temporal scales adopted by different cortical regions is seldom fully considered in the conventional EEG emotion recognition models. To solve this problem, first, the Deep-Wise Separable Convolution (DWSC) is modified to obtain Modified DWSC (MDWSC), and the multi- scale temporal features are extracted from each EEG channel with MDWSCs of different kernel sizes. Next, Coordinate Attention (CA) is introduced to utilize the fused multi-scale temporal feature to optimize the local spatial feature by reducing the heterogeneity existing in the temporal scale among cortical regions. Then, the Graph Convolution Network (GCN) and the Retention Transformer are introduced to grasp the global spatial information contained in the optimized local spatial feature and achieve gradual integration of information across time to capture the global temporal feature, respectively. Finally, the global spatial feature is fused with the integrated temporal feature for emotion recognition. Extensive experimental results under both subject- dependent and subject-independent scenarios on DEAP, DREAMER, and PhyMER datasets demonstrate that: (i) The proposed model outperforms State-Of-The-Art (SOTA) EEG-based emotion recognition baselines. (ii) All the key modules contribute to the performance enhancement of the proposed model. (iii) The proposed model can take full advantage of the complementarities among different features organically to enhance emotion recognition performance.