DAMSDAN: Distribution-Aware Multi-Source Domain Adaptation Network for Cross-Domain EEG-based Emotion Recognition

TL;DR

DAMSDAN tackles inter-subject variability in EEG-based emotion recognition by performing distribution-aware multi-source domain adaptation. It unifies marginal and conditional distribution alignment through a three-part architecture: a dual-encoder feature extractor, a marginal alignment module with prototype and adversarial components plus source weighting, and a conditional alignment module with dual pseudo-label collaboration and prototype-guided alignment. The method demonstrates strong cross-subject and cross-session performance on SEED, SEED-IV, and FACED, with extensive ablations confirming the necessity of each component and interpretability analyses highlighting informative EEG bands and brain regions. The approach offers a practical, scalable solution for cross-domain affective brain-computer interfaces and sets a foundation for robust, domain-aware EEG emotion recognition in real-world settings.

Abstract

Significant inter-individual variability limits the generalization of EEG-based emotion recognition under cross-domain settings. We address two core challenges in multi-source adaptation: (1) dynamically modeling distributional heterogeneity across sources and quantifying their relevance to a target to reduce negative transfer; and (2) achieving fine-grained semantic consistency to strengthen class discrimination. We propose a distribution-aware multi-source domain adaptation network (DAMSDAN). DAMSDAN integrates prototype-based constraints with adversarial learning to drive the encoder toward discriminative, domain-invariant emotion representations. A domain-aware source weighting strategy based on maximum mean discrepancy (MMD) dynamically estimates inter-domain shifts and reweights source contributions. In addition, a prototype-guided conditional alignment module with dual pseudo-label interaction enhances pseudo-label reliability and enables category-level, fine-grained alignment, mitigating noise propagation and semantic drift. Experiments on SEED and SEED-IV show average accuracies of 94.86\% and 79.78\% for cross-subject, and 95.12\% and 83.15\% for cross-session protocols. On the large-scale FACED dataset, DAMSDAN achieves 82.88\% (cross-subject). Extensive ablations and interpretability analyses corroborate the effectiveness of the proposed framework for cross-domain EEG-based emotion recognition.

Figures (9)

• Figure 1: Overall framework of the proposed DAMSDAN model.
• Figure 2: Architecture of the FE module.
• Figure 3: Illustration of the DASW strategy.
• Figure 4: Illustration of the PGCA strategy.
• Figure 5: Subject-wise classification accuracy under cross-subject leave-one-subject-out cross-validation on the SEED and SEED-IV datasets.