본문 바로가기
  • Home

Band-Attention BFE-Net for Subject-Independent EEG Emotion Recognition

  • Journal of The Korea Society of Computer and Information
  • Abbr : JKSCI
  • 2026, 31(3), pp.61~69
  • Publisher : The Korean Society Of Computer And Information
  • Research Area : Engineering > Computer Science
  • Received : January 21, 2026
  • Accepted : March 9, 2026
  • Published : March 31, 2026

Yujin Ji 1 Jungpyo Hong 1

1국립창원대학교

Accredited

ABSTRACT

Electroencephalogram (EEG) signals serve as a core input for brain–computer interface systems, and EEG-based emotion recognition has been widely studied. The Band Feature Extraction Network (BFE-Net) effectively extracts frequency band–specific features; however, it does not consider interactions between frequency bands. To address this limitation, this study proposes the Band-Attention BFE-Net (BA-BFE-Net), which incorporates a self-attention mechanism to explicitly learn dependencies among frequency bands. The proposed model captures non-linear relationships among five frequency bands and enables more effective feature fusion. Subject-independent experiments conducted on the SEED and SEED-IV datasets demonstrate that the proposed model improves classification performance by 1.19% and 2.04%, respectively, compared to the original BFE-Net.

Citation status

* References for papers published after 2024 are currently being built.