Annual Conference of KIPS (한국정보처리학회:학술대회논문집)

  • 2024.10a
  • /
  • Pages.573-576
  • /
  • 2024
  • /
  • 2005-0011(pISSN)
  • /
  • 2671-7298(eISSN)
Korea Information Processing Society (한국정보처리학회)
권호 표지
DOI QR코드

DOI QR Code

A Hybrid CNN-LSTM Approach for Effective Denoising of EEG Signals Contaminated by EOG Artifacts

  • Battulga Ulziisaikhan (Department of AI Convergence, Chonnam National University) ;
  • Nguyen Trong Nghia (Department of AI Convergence, Chonnam National University) ;
  • Soo-Hyung Kim (Department of AI Convergence, Chonnam National University)
  • Published : 2024.10.31
Download PDF
⟨ Previous Next ⟩

Abstract

Electroencephalography (EEG) signals are often contaminated with artifacts, particularly those from eye movements, recorded as electrooculography (EOG). Effective denoising methods are essential for accurate EEG analysis. In this paper, we compare different denoising approaches, focusing on both convolutional neural networks (CNNs) and recurrent neural networks (RNNs) for artifact removal. Through experiments, we found that CNNs excel in capturing spatial features, particularly in high-frequency EEG bands like Alpha and Beta, while RNNs are more effective at modeling temporal dependencies, particularly in lower-frequency bands like Delta and Theta. To leverage the strengths of both models, we propose a hybrid CNN-LSTM architecture. Our results show that the hybrid model achieves superior performance in denoising across all EEG frequency bands, with significant improvements in the Alpha and Beta bands. This approach provides a robust solution for denoising EEG signals contaminated with EOG artifacts, offering improved accuracy over standalone CNN or RNN models.

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-2024-RS-2024-00437718) supervised by IITP, and the National Research Foundation of Korea(NRF) grant funded by the Korea government(MSIT) (RS-2023-00219107).

References

  1. Hyvarinen A and Oja E 2000 Independent component analysis: algorithms and applications Neural Netw. 
  2. Wang G, Teng C, Li K, Zhang Z and Yan X 2016 The removal of EOG artifacts from EEG signals using independent component analysis and multivariate empirical mode decomposition IEEE J. Biomed. Health Inform.]
  3. Croft R J and Barry R J 2000 Removal of ocular artifact from the EEG: a review Clin. Neurophysiol. 30 5-19
  4. Cohen M X 2014 Analyzing Neural Time Series Data: Theory and Practice (Cambridge, MA: MIT Press).
  5. T. N. Nguyen, H. -J. Yang, B. -G. Kho, S. -R. Kang and S. -H. Kim, "Explainable Deep Contrastive Federated Learning System for Early Prediction of Clinical Status in Intensive Care Unit," in IEEE Access, vol. 12, pp. 117176-117202, 2024, doi: 10.1109/ACCESS.2024.3447759.
  6. Cormen, T. H., Leiserson, C. E., Rivest, R. L., Stein, C., "Introduction to Algorithms", Cambridge, MIT Press, 2009.
  7. Zhang, H.; Zhao, M.; Wei, C.; Mantini, D.; Li, Z.; Liu, Q. EEGdenoiseNet: A benchmark dataset for end-to-end deep learning solutions of EEG denoising. arXiv 2020, arXiv:2009.11662.
  8. Wang G, Teng C, Li K, Zhang Z and Yan X 2016 Theremoval of EOG artifacts from EEG signals using independent component analysis and multivariate empirical mode decomposition IEEE J. Biomed. Health Inform.
  9. Zhang, Y., et al. "EEG Signal Denoising Based on Wavelet Transform and Improved Threshold Function." Cognitive Neurodynamics, 2018.