Electroencephalography (EEG)-based driver-drowsiness detection is hindered by cross-subject variability and the resource cost of multi-channel models. This article presents Lite dynamic vision with explicit-adaptive implicit matching knowledge distillation (Lite-DV-EAIM-KD), a lightweight, subject-adaptive, and interpretable framework for binary drowsiness recognition. Thirty-channel EEG is decomposed into six wavelet components and represented as raw and Euclidean-aligned views. Each view employs a shared convolutional trunk with component-specific normalization, temporal attention, and component attention. A larger dual-view teacher provides offline supervision during unlabeled target-subject adaptation, and only the compact two-view student is retained for inference. Leave-one-subject-out experiments on 2,022 balanced three-second EEG epochs from 11 subjects yield 83.400% mean accuracy, 83.330% macro-average F1 score (macro-F1), and 89.580% mean area under the curve (AUC). Compared with the non-distilled shared-component student, the proposed method gains 4.37 percentage points in accuracy while retaining 4,112 deployed parameters. It reduces parameters by 82.600% and median batch-one graphics processing unit (GPU) latency by 74.900% relative to the teacher. Attribution-guided occlusion further supports the faithfulness of the learned frequency and channel evidence. These results show that compact multi-component EEG modeling can preserve cross-subject discrimination when coupled with unlabeled target adaptation and explicit decision evidence.
References
[1] Al-Shargie, F., Tariq, U., Hassanin, O., Mir, H., Babiloni, F., Al-Nashash, H. (2019) Brain connectivity analysis under semantic vigilance and enhanced mental states. Brain Sciences, 9(12), 363.
[2] He, L., Zhang, L., Sun, Q., Lin, X. (2024) A generative adaptive convolutional neural network with attention mechanism for driver fatigue detection with class-imbalanced and insufficient data. Behavioural Brain Research, 464, 114898.
[3] Luo, Y., Liu, W., Li, H., Lu, Y., Lu, B. L. (2024) A cross-scenario and cross-subject domain adaptation method for driving fatigue detection. Journal of Neural Engineering, 21(4), 046004.
[4] Yuan, L., Cui, J., Li, R., Zheng, Z., Siyal, M. Y., Yi, Z. (2024) Entropy-guided robust feature domain adaptation for electroencephalogram-based cross-dataset drowsiness recognition. Engineering Applications of Artificial Intelligence, 137, 109153.
[5] He, X., Li, H., Yu, P., Wu, H., Chen, B. (2025) DP-MP: a novel cross-subject fatigue detection framework with DANN-based prototypical representation and mix-up pairwise learning. Journal of Neural Engineering, 22(2), 026049.
[6] Lawhern, V. J., Solon, A. J., Waytowich, N. R., Gordon, S. M., Hung, C. P., Lance, B. J. (2018) EEGNet: a compact convolutional neural network for EEG-based brain – computer interfaces. Journal of Neural Engineering, 15(5), 056013.
[7] Feng, W., Wang, X., Xie, J., Liu, W., Qiao, Y., Liu, G. (2024) Real-time EEG-based driver drowsiness detection based on convolutional neural network with gumbel-softmax trick. IEEE Sensors Journal, 25(1), 1860-1871.
[8] Zayed, A., Trabes, E., Tarrillo, J., Ben Khalifa, K., Valderrama, C. (2025) Efficient embedded system for drowsiness detection based on EEG signals: features extraction and hardware acceleration. Electronics, 14(3), 404.
[9] Pillalamarri, R., Shanmugam, U. (2026) Enhancing cybersecurity via an emotion aware framework: leveraging EEG and speech signal fusion through correlation-based networks. Journal on Information Security, 2026(1), 11.
[10] Cui, J., Lan, Z., Sourina, O., Müller-Wittig, W. (2022) EEG-based cross-subject driver drowsiness recognition with an interpretable convolutional neural network. IEEE Transactions on Neural Networks and Learning Systems, 34(10), 7921-7933.
[11] Cui, J., Lan, Z., Liu, Y., Li, R., Li, F., Sourina, O., Müller-Wittig, W. (2022) A compact and interpretable convolutional neural network for cross-subject driver drowsiness detection from single-channel EEG. Methods, 202, 173-184.
[12] Feng, X., Guo, Z., Kwong, S. (2025) ID3RSNet: cross-subject driver drowsiness detection from raw single-channel EEG with an interpretable residual shrinkage network. Frontiers in Neuroscience, 18, 1508747.
[13] He, H., Wu, D. (2019) Transfer learning for brain – computer interfaces: a Euclidean space data alignment approach. IEEE Transactions on Biomedical Engineering, 67(2), 399-410.
[14] Junqueira, B., Aristimunha, B., Chevallier, S., de Camargo, R. Y. (2024) A systematic evaluation of Euclidean alignment with deep learning for EEG decoding. Journal of Neural Engineering, 21(3), 036038.
[15] Cowley, B. R., Stan, P. L., Pillow, J. W., Smith, M. A. (2026) Compact deep neural network models of the visual cortex. Nature, 652(8111), 947-954.
Share and Cite
Hou, Q. (2026) A Lightweight Subject-adaptive EEG Network for Driver Drowsiness Detection. Scientific Research Bulletin, 3(2), 90-99. https://doi.org/10.71052/srb2024/DGII1415