The mitigation of heterogeneity in temporal scale among different cortical regions for EEG emotion recognition

Zhangyong Xu, Ning Chen, Guangqiang Li, Jing Li, Hongqing Zhu, Zhiying Zhu

Abstract: Highlights•Neuroscience prior knowledge is introduced to design EEG emotion recognition model.•DWSC is modified to capture the multi-scale local temporal feature in EEG.•Coordinate attention is adopted to reduce temporal heterogeneity in cortical regions.•Retention Transformer is introduced to capture the global temporal feature in EEG.•It outperforms SOTA models in both 2 and 4 class emotion classification tasks.