Driver fatigue is a major cause of traffic accidents. Electroencephalography (EEG), which directly reflects neural activity, has been widely used for fatigue detection. However, existing methods still suffer from limited interpretability of features, high subjectivity in fatigue labeling, and insufficient capability in modeling temporal dependencies. To address these issues, this study proposes a fatigue detection method based on EEG band-ratio features and eye-movement-derived weak supervision labels. A deep learning model, namely the stacked long short-term memory attention network (stacked LSTM attention network, SLAN), is designed to capture temporal dependencies in EEG signals. After preprocessing, five EEG band-ratio features are extracted, and a dual-threshold pupil diameter strategy is used to construct three-level fatigue labels for weakly supervised learning. An attention mechanism is further introduced to enhance temporal feature modeling and improve interpretability by focusing on critical time segments. Experiments on a self-collected dataset of 20 subjects demonstrate that the proposed model achieves an accuracy of approximately 97.0% in the three-class fatigue classification task, outperforming EEG-based spatio-temporal convolutional neural network (ESTCNN), EEGNet, and Interpretable convolutional neural network (CNN) across all evaluation metrics. The proposed approach effectively improves temporal representation and interpretability of fatigue states and reveals dynamic temporal patterns of EEG activity under fatigue conditions, providing a foundation for continuous driver state monitoring and adaptive warning systems in intelligent cockpit applications.
References
[1] Nguyen, K., Dunbar, C., Guyett, A., Bickley, K., Nguyen, D. P., Reynolds, A. C., Vakulin, A. (2025) Poorer objective but not subjective driving performance in drivers vulnerable to sleep loss effects during extended wake. Journal of Sleep Research, 34(4), e14455.
[2] Sikander, G., Anwar, S. (2018) Driver fatigue detection systems: a review. IEEE Transactions on Intelligent Transportation Systems, 20(6), 2339-2352.
[3] Hassan, O. F., Ibrahim, A. F., Gomaa, A., Makhlouf, M. A., Hafiz, B. (2025) Real-time driver drowsiness detection using transformer architectures: a novel deep learning approach. Scientific Reports, 15(1), 17493.
[4] Mora, H. J., Echaveguren, T. B., Pino, E. J. (2026) Advanced forecasting of driver drowsiness events: non-intrusive data and multimodal BiLSTM-based modeling. Biomedical Signal Processing and Control,119, 109793.
[5] Skaiky, A. A., Ayad, H. (2025) Deep learning approaches for fatigue detection: a traditional review of models, datasets, and applications. Al-Farahidi Expert Systems Journal, 1(2), 10.
[6] AlArnaout, Z., Zaki, C., Kotb, Y., AlAkkoumi, M., Mostafa, N. (2025) Exploiting heart rate variability for driver drowsiness detection using wearable sensors and machine learning. Scientific Reports, 15(1), 24898.
[7] Fowler, A., Harvey, C., Wilson, M. L., Sharples, S. (2026) Using wearable measures to infer moments in workload from Electrodermal Activity and individual workload from Heart Rate Variability during a simulated railway signalling task. Applied Ergonomics, 133, 104710.
[8] Wang, Z., Du, X., Jiang, C., Sun, J. (2025) Research on the prediction of driver fatigue degree based on EEG signals. Sensors, 25(23), 7316.
[9] Kalogeropoulos, C., Theofilatos, K., Mavroudi, S. (2026) From neurons to networks: a holistic review of electroencephalography (EEG) from neurophysiological foundations to AI techniques. Signals, 7(1), 17.
[10] Li, A., Wang, Z., Xu, T., Zhou, T., Zhao, X., Hu, H., Van Hulle, M. M. (2026) A cross-subject band-power complexity metric for detecting mental fatigue through EEG. Brain Sciences, 16(2), 199.
[11] Yang, B., Ding, Y., Cui, C., Guo, T. (2025) A hybrid kernel principal component analysis and long short-term memory approach for EEG signal-based fatigue evaluation. Alexandria Engineering Journal, 131, 209-217.
[12] Zhang, Y., Xu, X., Du, Y., Zhang, N. (2025) TMU-net: a transformer-based multimodal framework with uncertainty quantification for driver fatigue detection. Sensors, 25(17), 5364.
[13] Bello, I., Sehnan, M., Dang, W., Banzi, J. F., Aminu, S. A., Gao, Z. (2026) A frequency-attentive spatio-temporal convolutional neural network with weighted feature fusion for EEG-based driver fatigue detection. Biomedical Signal Processing and Control, 113, 108979.
[14] Kim, S., Wisanggeni, I., Ros, R., Hussein, R. (2020) Detecting fatigue driving through PERCLOS: a review. International Journal of Image Processing (IJIP), 14(1), 1-7.
[15] Quddus, A., Zandi, A. S., Prest, L., Comeau, F. J. (2021) Using long short term memory and convolutional neural networks for driver drowsiness detection. Accident Analysis & Prevention, 156, 106107.
Share and Cite
Xu, F., Yu, P., Guo, H. (2026) EEG-based Driver Fatigue Detection with Eye-tracking Guided Weak Supervision. Scientific Research Bulletin, 3(2), 27-37. https://doi.org/10.71052/srb2024/AAFR1341