A Fine-grained Hemispheric Asymmetry Network for accurate and interpretable EEG-based emotion classification

• DOI: 10.1016/j.neunet.2025.107127
• Publication Date: 2025-01-09T00:00:51Z
• Authors: Ruofan Yan; Na Lu; Yuxuan Yan; Xu Niu; Jibin Wu

In this work, we propose a Fine-grained Hemispheric Asymmetry Network (FG-HANet), an end-to-end deep learning model that leverages hemispheric asymmetry features within 2-Hz narrow frequency bands for accurate and interpretable emotion classification over raw EEG data. particular, the FG-HANet extracts not only from original inputs but also their mirrored versions, applies Finite Impulse Response (FIR) filters at granularity as fine to acquire fine-grained spectral information. Furthermore, guarantee sufficient attention features, tailor three-stage training pipeline further boost its performance. We conduct extensive evaluations on two public datasets, SEED SEED-IV, experimental results well demonstrate superior performance of proposed FG-HANet, i.e. 97.11% 85.70% accuracy, respectively, building new state-of-the-art. Our reveal dominance under different emotional states hemisphere in individuals. These align with previous findings neuroscience provide insights into underlying generation mechanisms.