Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A novel Fourier Adjacency Transformer for advanced EEG emotion recognition

Jinfeng Wang, Yanhao Huang, Sifan Song, Boqian Wang, Jionglong Su, Jiaman Ding

TL;DR

This work tackles the challenges of EEG emotion recognition by integrating Fourier-based periodic modeling with graph-inspired attention. The authors introduce the Fourier Analytic Linear (FAL) layer to decouple periodic and aperiodic EEG components, preserving linearity for the aperiodic path, and combine it with Fourier Adjacent Attention (FAA) to learn both universal and sample-specific inter-channel connectivities. Replacing standard self-attention with FAA yields the Fourier Adjacent Transformer (FAT), which processes Differential Entropy features via patch and positional embeddings. FAT achieves state-of-the-art results on SEED family datasets and DEAP, with improvements up to approximately 6.5% in several tasks, and demonstrates strong generalization and interpretability through adjacency-based analysis of inter-channel relationships.

Abstract

EEG emotion recognition faces significant hurdles due to noise interference, signal nonstationarity, and the inherent complexity of brain activity which make accurately emotion classification. In this study, we present the Fourier Adjacency Transformer, a novel framework that seamlessly integrates Fourier-based periodic analysis with graph-driven structural modeling. Our method first leverages novel Fourier-inspired modules to extract periodic features from embedded EEG signals, effectively decoupling them from aperiodic components. Subsequently, we employ an adjacency attention scheme to reinforce universal inter-channel correlation patterns, coupling these patterns with their sample-based counterparts. Empirical evaluations on SEED and DEAP datasets demonstrate that our method surpasses existing state-of-the-art techniques, achieving an improvement of approximately 6.5% in recognition accuracy. By unifying periodicity and structural insights, this framework offers a promising direction for future research in EEG emotion analysis.

A novel Fourier Adjacency Transformer for advanced EEG emotion recognition

TL;DR

This work tackles the challenges of EEG emotion recognition by integrating Fourier-based periodic modeling with graph-inspired attention. The authors introduce the Fourier Analytic Linear (FAL) layer to decouple periodic and aperiodic EEG components, preserving linearity for the aperiodic path, and combine it with Fourier Adjacent Attention (FAA) to learn both universal and sample-specific inter-channel connectivities. Replacing standard self-attention with FAA yields the Fourier Adjacent Transformer (FAT), which processes Differential Entropy features via patch and positional embeddings. FAT achieves state-of-the-art results on SEED family datasets and DEAP, with improvements up to approximately 6.5% in several tasks, and demonstrates strong generalization and interpretability through adjacency-based analysis of inter-channel relationships.

Abstract

EEG emotion recognition faces significant hurdles due to noise interference, signal nonstationarity, and the inherent complexity of brain activity which make accurately emotion classification. In this study, we present the Fourier Adjacency Transformer, a novel framework that seamlessly integrates Fourier-based periodic analysis with graph-driven structural modeling. Our method first leverages novel Fourier-inspired modules to extract periodic features from embedded EEG signals, effectively decoupling them from aperiodic components. Subsequently, we employ an adjacency attention scheme to reinforce universal inter-channel correlation patterns, coupling these patterns with their sample-based counterparts. Empirical evaluations on SEED and DEAP datasets demonstrate that our method surpasses existing state-of-the-art techniques, achieving an improvement of approximately 6.5% in recognition accuracy. By unifying periodicity and structural insights, this framework offers a promising direction for future research in EEG emotion analysis.

Paper Structure

This paper contains 15 sections, 8 equations, 3 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Methods
  3. Fourier Analytic Linear
  4. Fourier Adjacent Attention
  5. Fourier Adjacent Transformer
  6. Validation scheme
  7. Results and discussions
  8. Implementation details
  9. Datasets
  10. Experimental settings
  11. Recognition capability
  12. Generalization capability
  13. Ablation study
  14. Visualization
  15. Conclusion

Figures (3)

  • Figure 1: Illustrations of MLP Layer vs. FAN Layer vs. FAL Layer.
  • Figure 2: The overview of the FAT and FAA.
  • Figure 3: The top 15 correlations in each of the two adjacency matrices.