Journal of Biomedical Engineering Research (대한의용생체공학회:의공학회지)

The Korean Society of Medical and Biological Engineering (대한의용생체공학회)
권호 표지
DOI QR코드

DOI QR Code

Diffusion-Based Data Augmentation for Few-Channel EEG Emotion Recognition Using Class-Conditional DDPM

소수 채널 EEG 감정 인식을 위한 클래스 조건부 확산 모델 기반 데이터 증강 기법

  • Kihyeon Joo (Department of Electrical and Computer Engineering, Inha University) ;
  • ChaeEun Woo (Department of Electrical and Computer Engineering, Inha University) ;
  • Byung Hyung Kim (Department of Artificial Intelligence, Inha University)
  • 주기현 (인하대학교 전기컴퓨터공학과) ;
  • 우채은 (인하대학교 전기컴퓨터공학과) ;
  • 김병형 (인하대학교 인공지능공학과)
  • Received : 2025.11.18
  • Accepted : 2026.01.12
  • Published : 2026.02.28
Download PDF
⟨ Previous Next ⟩

Abstract

This study proposes a diffusion-based data augmentation framework optimized for few-channel EEG emotion recognition, addressing the intrinsic data scarcity and information loss inherent in low-cost, portable EEG systems. Using the SEED dataset, we utilized 30-dimensional differential entropy (DE) features extracted from six frontal channels. To effectively model these low-dimensional tabular features, we designed a class-conditional ϵ-prediction Denoising Diffusion Probabilistic Model (DDPM) incorporating a FiLM-ResMLP architecture. Furthermore, we introduced a novel Mean-Guided Guidance Sampling mechanism to explicitly steer generated samples toward class-consistent statistical distributions, thereby enhancing training stability. Experimental results demonstrate that the proposed FiLM-ResMLP based Conditional DDPM achieves a classification accuracy of 57.87%, representing a significant improvement of 4.85%p over the baseline (53.02%). The proposed method consistently outperforms state-of-the-art generative baselines, including Conditional DDPM (FiLM-UNet), Conditional WGAN and VAED2GAN. Notably, our analysis reveals that while the conventional UNet architecture tends to overfit in the low-dimensional feature space, the proposed ResMLP structure exhibits superior robustness and statistical compatibility with tabular EEG data. By demonstrating that data augmentation can effectively compensate for information loss caused by channel reduction, this work presents a viable methodology for overcoming the performance limitations of practical few-channel BCI systems.

Keywords

Acknowledgement

본 연구는 정부(과학기술정보통신부)의 재원으로 한국연구재단의 지원(No. 2021R1C1C2012437), 그리고 인하대학교 연구비의 지원을 받아 수행되었음.

References

  1. Koelstra S, Muhl C, Soleymani M, Lee JS, Yazdani A, Ebrahimi T. DEAP: A Database for Emotion Analysis;Using Physiological Signals. IEEE Transactions on Affective Computing. 2012,3(1):18-31. https://doi.org/10.1109/T-AFFC.2011.15
  2. Duan RN, Zhu JY, Lu BL. Differential entropy feature for EEG-based emotion classification. 2013 6th International IEEE/EMBS Conference on Neural Engineering (NER). 2013;81-84.
  3. Zheng WL, Liu W, Lu Y, Lu BL, Cichocki A. EmotionMeter: A Multimodal Framework for Recognizing Human Emotions. IEEE Transactions on Cybernetics. 2019;49(3):1110-1122. https://doi.org/10.1109/TCYB.6221036
  4. Soleymani M, Lichtenauer J, Pun T, Pantic M. A Multimodal Database for Affect Recognition and Implicit Tagging. IEEE Transactions on Affective Computing. 2012;3(1):42-55. https://doi.org/10.1109/T-AFFC.2011.25
  5. Miranda-Correa JA, Abadi MK, Sebe N, Patras I. AMIGOS: A Dataset for Affect, Personality and Mood Research on Individuals and Groups. IEEE Transactions on Affective Computing. 2021;12(2):479-493. https://doi.org/10.1109/TAFFC.2018.2884461
  6. AI-Hadithy SS, Abdalkafor AS, AI-Khateeb B. Emotion recognition in EEG Signals: Deep and machine learning approaches, challenges, and future directions. Computers in Biology and Medicine. 2025;196:110713. https://doi.org/10.1016/j.compbiomed.2025.110713
  7. Dabel MA. A Review on Data Scarcity in EEG-based Emotion Recognition. International Journal of Computer Science and Network Security. 2025;25(4):238-248. https://doi.org/10.22937/IJCSNS.2025.25.4.24
  8. Woo CE, Choi HS, Kim BH. Valence-Arousal Regression Model Using EEG with Multi-Output Prediction. Journal of Biomedical Engineering Research. 2024;45(5):279-285. https://doi.org/10.9718/JBER.2024.45.5.279
  9. Yuvaraj R, Thagavel P, Thomas J, Fogarty J, Ali F. Comprehensive Analysis of Feature Extraction Methods for Emotion Recognition from Multichannel EEG Recordings. Sensors. 2023;23(2):915. https://doi.org/10.3390/s23020915
  10. Lu W, Liu H, Ma H, Tan TP, Xia L. Hybrid transfer learning strategy for cross-subject EEG emotion recognition. Frontiers in Human Neuroscience. 2023;17:1280241.
  11. Wu X, Ju X, Dai S, Li X, Li M. Multi-source domain adaptation for EEG emotion recognition based on inter-domain sample hybridization. Frontiers in Human Neuroscience. 2024;18:1464431. https://doi.org/10.3389/fnhum.2024.1464431
  12. Luo Y, Lu BL. EEG Data Augmentation for Emotion Recognition Using a Conditional Wasserstein GAN. 2018 40th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC). 2018;2535-2538.
  13. Bao G, Yan B, Tong L, Shu J, Wang L, Yang K, Zeng Y. Data Augmentation for EEG-Based Emotion Recognition Using Generative Adversarial Networks. Frontiers in Computational Neuroscience. 2021;15. https://doi.org/10.3389/fncom.2021.723843
  14. Zhao YD, Liu YK, Zheng WL, Lu BL. EEG Data Augmentation for Emotion Recognition Using Diffusion Model. 2024 46th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC). 2024;1-4.
  15. Chen M, Gui Y, Su Y, Zhu Y, Luo G, Yang Y. Improving EEG Classification Through Randomly Reassembling Original and Generated Data with Transformer-based Diffusion Models. arXiv. arxiv:2407.20253. 2024.
  16. Huang J, Li M, Chen W. SAD-VER: A Self-supervised, Diffusion probabilistic model-based data augmentation framework for Visual-stimulus EEG Recognition. Advanced Engineering Informatics. 2025;65:103298. https://doi.org/10.1016/j.aei.2025.103298
  17. Topic A, Russo M, Stella M, Saric M. Emotion Recognition Using a Reduced Set of EEG Channels Based on Holographic Feature Maps. Sensors. 2022;22(9):3248. https://doi.org/10.3390/s22093248
  18. Xiong F, Fan M, Yang X, Wang C, Zhou J. Research on emotion recognition using sparse EEG channels and cross-subject modeling based on CNN-KAN-F2CA model. PLOS ONE. 2025;20(5):e0322583. https://doi.org/10.1371/journal.pone.0322583
  19. Liu F, Yang P, Shu Y, Liu N, Sheng J, Luo J, Wang X, Liu YJ. Emotion Recognition From Few-Channel EEG Signals by Integrating Deep Feature Aggregation and Transfer Learning. IEEE Transactions on Affective Computing. 2024;15(3): 1315-1330. https://doi.org/10.1109/TAFFC.2023.3336531
  20. Valderrama CE, Sheoran A. Identifying relevant EEG channels for subject-independent emotion recognition using attention network layers. Frontiers in Psychiatry. 2025;16. https://doi.org/10.3389/fpsyt.2025.1494369
  21. Ho J, Jain A, Abbeel P. Denoising Diffusion Probabilistic Models. Advances in Neural Information Processing Systems 33. 2020;6840-6851.
  22. https://bcmi.sjtu.edu.cn/home/seed/seed.html. Accessed on 30 Oct 2025.
  23. Perez E, Strub F, de Vries H, Dumouling V, Courville A. FiLM: visual reasoning with a general conditioning layer. Proceedings of the Thirty-Second AAAI Conference on Artificial Intelligence and Thirtieth Innovative Applications of Artificial Intelligence Conference and Eighth AAAI Symposium on Educational Advances in Artificial Intelligence. 2018;3942-3951.
  24. Ho J, Salimans T. Classifier-Free Diffusion Guidance. arXiv. arXiv:2207.12598. 2022.