Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

EEG-based Graph-guided Domain Adaptation for Robust Cross-Session Emotion Recognition

Maryam Mirzaei, Farzaneh Shayegh, Hamed Narimani

TL;DR

This work tackles cross-session variability in EEG-based emotion recognition by introducing EGDA, a graph-guided domain adaptation framework that jointly aligns marginal and conditional distributions while preserving intrinsic data geometry. It learns a shared subspace via a projection matrix $A$ and builds an adaptive similarity graph $S$ to maintain neighborhood structure, using iterative pseudo-label updates to refine class-level alignment. On SEED-IV, EGDA achieves strong cross-session performance (e.g., accuracies around 81–83% across transfer tasks) and outperforms several baselines, with Gamma frequency and fronto-parietal regions identified as most discriminative. The approach offers a robust, scalable method for cross-session emotion recognition with practical implications for reliable human–machine interaction across sessions.

Abstract

Accurate recognition of human emotional states is critical for effective human-machine interaction. Electroencephalography (EEG) offers a reliable source for emotion recognition due to its high temporal resolution and its direct reflection of neural activity. Nevertheless, variations across recording sessions present a major challenge for model generalization. To address this issue, we propose EGDA, a framework that reduces cross-session discrepancies by jointly aligning the global (marginal) and class-specific (conditional) distributions, while preserving the intrinsic structure of EEG data through graph regularization. Experimental results on the SEED-IV dataset demonstrate that EGDA achieves robust cross-session performance, obtaining accuracies of 81.22%, 80.15%, and 83.27% across three transfer tasks, and surpassing several baseline methods. Furthermore, the analysis highlights the Gamma frequency band as the most discriminative and identifies the central-parietal and prefrontal brain regions as critical for reliable emotion recognition.

EEG-based Graph-guided Domain Adaptation for Robust Cross-Session Emotion Recognition

TL;DR

This work tackles cross-session variability in EEG-based emotion recognition by introducing EGDA, a graph-guided domain adaptation framework that jointly aligns marginal and conditional distributions while preserving intrinsic data geometry. It learns a shared subspace via a projection matrix and builds an adaptive similarity graph to maintain neighborhood structure, using iterative pseudo-label updates to refine class-level alignment. On SEED-IV, EGDA achieves strong cross-session performance (e.g., accuracies around 81–83% across transfer tasks) and outperforms several baselines, with Gamma frequency and fronto-parietal regions identified as most discriminative. The approach offers a robust, scalable method for cross-session emotion recognition with practical implications for reliable human–machine interaction across sessions.

Abstract

Accurate recognition of human emotional states is critical for effective human-machine interaction. Electroencephalography (EEG) offers a reliable source for emotion recognition due to its high temporal resolution and its direct reflection of neural activity. Nevertheless, variations across recording sessions present a major challenge for model generalization. To address this issue, we propose EGDA, a framework that reduces cross-session discrepancies by jointly aligning the global (marginal) and class-specific (conditional) distributions, while preserving the intrinsic structure of EEG data through graph regularization. Experimental results on the SEED-IV dataset demonstrate that EGDA achieves robust cross-session performance, obtaining accuracies of 81.22%, 80.15%, and 83.27% across three transfer tasks, and surpassing several baseline methods. Furthermore, the analysis highlights the Gamma frequency band as the most discriminative and identifies the central-parietal and prefrontal brain regions as critical for reliable emotion recognition.
Paper Structure (15 sections, 30 equations, 7 figures, 6 tables, 1 algorithm)

This paper contains 15 sections, 30 equations, 7 figures, 6 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. SEED-IV–Based emotion recognition methods
  4. Transfer Learning Approaches
  5. PROPOSED METHOD
  6. Preliminary
  7. Problem Statement
  8. Model Formulation
  9. Model Optimization
  10. Experiments
  11. Dataset Description
  12. Overall Cross-Session Performance
  13. Emotion-Specific Recognition
  14. Critical Frequency Bands and Channels Identification
  15. Conclusion

Figures (7)

  • Figure 1: Accuracies of different models for cross-session EEG emotion recognition
  • Figure 2: The average confusion graph of for cross-session EEG emotion recognition.
  • Figure 3: Visualization of EEG feature importance with respect to frequency bands and channels.
  • Figure 4: Average importance of EEG frequency bands obtained by for cross-session emotion recognition
  • Figure 5: Average EEG channel importance obtained by for cross-session emotion recognition.
  • ...and 2 more figures