Region-aware Spatiotemporal Modeling with Collaborative Domain Generalization for Cross-Subject EEG Emotion Recognition

TL;DR

This work tackles cross-subject EEG emotion recognition under substantial inter-subject variability by introducing RSM-CoDG, a neuroscience-informed framework that jointly learns region-aware spatial representations, multi-scale temporal dynamics, and collaborative domain generalization. The Region-aware Graph Representation Module imposes functional brain priors via six regions, the Multi-Scale Temporal Transformer captures local and global temporal dependencies, and the Collaboratively Optimized DG Strategy enforces distribution, attention, and structure alignment without target-domain data. The approach achieves state-of-the-art cross-subject performance on SEED, SEED-IV, and SEED-V, with robust subject-level metrics and interpretable activations aligned to neurophysiological patterns. While effective, the model incurs training-time overhead due to multi-loss optimization, suggesting avenues for efficiency-focused future work with lightweight regularization strategies.

Abstract

Cross-subject EEG-based emotion recognition (EER) remains challenging due to strong inter-subject variability, which induces substantial distribution shifts in EEG signals, as well as the high complexity of emotion-related neural representations in both spatial organization and temporal evolution. Existing approaches typically improve spatial modeling, temporal modeling, or generalization strategies in isolation, which limits their ability to align representations across subjects while capturing multi-scale dynamics and suppressing subject-specific bias within a unified framework. To address these gaps, we propose a Region-aware Spatiotemporal Modeling framework with Collaborative Domain Generalization (RSM-CoDG) for cross-subject EEG emotion recognition. RSM-CoDG incorporates neuroscience priors derived from functional brain region partitioning to construct region-level spatial representations, thereby improving cross-subject comparability. It also employs multi-scale temporal modeling to characterize the dynamic evolution of emotion-evoked neural activity. In addition, the framework employs a collaborative domain generalization strategy, incorporating multidimensional constraints to reduce subject-specific bias in a fully unseen target subject setting, which enhances the generalization to unknown individuals. Extensive experimental results on SEED series datasets demonstrate that RSM-CoDG consistently outperforms existing competing methods, providing an effective approach for improving robustness. The source code is available at https://github.com/RyanLi-X/RSM-CoDG.

Figures (11)

• Figure 1: Conceptual comparison of domain adaptation (DA) and domain generalization (DG) in cross-subject EEG emotion recognition. DA aligns source and target feature distributions using unlabeled target data during training, whereas DG learns a subject-invariant representation from multiple source subjects and generalizes directly to unseen targets without any target-domain data.
• Figure 2: Overall framework of the RSM-CoDG Model.
• Figure 3: Network structure diagram of RGRM.
• Figure 4: Illustration of the 62 EEG electrodes partitioned into six functional brain regions. Electrodes sharing the same color belong to the same brain area.
• Figure 5: Cross-subject experimental results for every participant in the SEED dataset.