Annual Conference of KIPS (한국정보처리학회:학술대회논문집)

  • 2025.05a
  • /
  • Pages.516-519
  • /
  • 2025
  • /
  • 2005-0011(pISSN)
  • /
  • 2671-7298(eISSN)
Korea Information Processing Society (한국정보처리학회)
권호 표지
DOI QR코드

DOI QR Code

Bandwise Feature Learning and Stacked Ensembling for EEG Sleep Stage Classification

EEG 수면 단계 분류를 위한 대역별 특징 학습 및 스택 앙상블 기법

  • Battulga Ulziisaikhan (Dept. of Artificial Intelligence Convergence Chonnam National University) ;
  • Soo-Hyung Kim (Dept. of Artificial Intelligence Convergence Chonnam National University) ;
  • Aera Kim (Dept. of Artificial Intelligence Convergence Chonnam National University)
  • 바툴가 (인공지능융합학과, 전남대학교 대학원) ;
  • 김수형 (인공지능융합학과, 전남대학교 대학원) ;
  • 김애라 (인공지능융합학과, 전남대학교 대학원)
  • Published : 2025.05.29
Download PDF
⟨ Previous Next ⟩

Abstract

We propose a Bandwise CNN ensemble framework for EEG-based sleep stage classification. The EEG signals were first decomposed into five standard frequency bands (Delta, Theta, Alpha, Beta, Gamma). For each band, an individual CNN was trained to capture band-specific features. The softmax outputs from these CNNs were then combined using a multilayer perceptron (MLP) meta-learner, enabling the model to make final predictions based on all band-level information. Compared to conventional full-signal models such as LSTM, CNN-LSTM, and Transformer, our ensemble approach achieved superior performance, reaching 92.23% accuracy on the Sleep-EDF dataset. Notably, the method demonstrated improved generalization to minority classes. These results highlight that frequency-specific modeling with CNNs can be more effective than sequential models for EEG-based sleep stage analysis.

Keywords

Acknowledgement

This work was supported by the Institute of Information & Communications Technology Planning & Evaluation (IITP) under the Artificial Intelligence Convergence Innovation Human Resources Development (IITP-2023-RS-2023-00256629) grant funded by the Korea government (MSIT), the ITRC (Information Technology Research Center) support program IITP-2025-RS-2024-00437718) supervised by IITP, the National Research Foundation of Korea(NRF) grant funded by the Korea government(MSIT) (RS-2023-00219107).

References

  1. B. Sharma, M., Tiwari, J., & Acharya, U. R. (2021). Automatic sleep-stage scoring in healthy and sleep disorder patients using optimal wavelet filter bank technique with EEG signals. International Journal of Environmental Research and Public Health, 18(6), 3087
  2. Voss, U., Holzmann, R., Hobson, A., Paulus, W., Koppehele-Gossel, J., Klimke, A., & Nitsche, M. A. (2014). Induction of self-awareness in dreams through frontal low current stimulation of gamma activity. Nature Neuroscience, 17(6), 810–812
  3. Hsu, Y., Yang, Y. C., Wang, J., & Hsu, C. (2013). Automatic sleep stage recurrent neural classifier using energy features of EEG signals. Neurocomputing, 104, 105–114.
  4. Hassan, A. R., & Bhuiyan, M. I. (2016). A decision support system for automatic sleep staging from EEG signals using tunable Q-factor wavelet transform and spectral features. Journal of Neuroscience Methods, 271, 107–118.
  5. Hassan, A. R., & Bhuiyan, M. I. (2017). Automated identification of sleep states from EEG signals by means of ensemble empirical mode decomposition and random under sampling boosting. Computer Methods and Programs in Biomedicine, 140, 201–210.
  6. Aboalayon, K. A. I., Faezipour, M., Almuhammadi, W., & Moslehpour, S. (2016). Sleep stage classification using EEG signal analysis: A comprehensive survey and new investigation. Entropy, 18(9), 272
  7. Memar, P., & Faradji, F. (2018). A novel multi-class EEG-based sleep stage classification system. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 26, 84–95.
  8. Acharya, U. R., Bhat, S., Faust, O., Adeli, H., Chua, E. C. P., Lim, W. J. E., & Koh, J. E. W. (2015). Nonlinear dynamics measures for automated EEG-based sleep stage detection. European Neurology, 74(5–6), 268–287