Driver Drowsiness Detection: Comparative Analysis of BiLSTM and CNN+BiLSTM Architectures

• Luma T. L. de Souza IFES
• Thiago M. Paixão IFES
• Richard J. M. G. Tello IFES

DOI: https://doi.org/10.5753/eries.2025.16019

Abstract

Driver drowsiness can negatively affect a person’s ability to stay alert, compromising not only their own safety but also the safety of others. In this work, we propose a comparison between a binary CNN+BiLSTM model and a BiLSTM model to predict whether an individual is alert or not. Using the UTA-RLLD dataset, which contains videos of individuals actually experiencing drowsiness, we process the positions of the eyes and mouth, as well as the distance between the chin and the nose, transforming these features into vectors that allow the model to capture the spatial information of each frame. A Bidirectional Long Short-Term Memory (BiLSTM) network is employed to capture the temporal dynamics across frames, including gradual changes in eye closure, yawning, and head movements. The experimental results show that the CNN+BiLSTM model achieves higher accuracy on the test dataset (77.59%) compared to the model using only BiLSTM layers (70.69%), demonstrating the advantage of integrating convolutional layers with BiLSTM.

References

ABRAMET (2020). Problemas na saúde de motoristas causaram mais de 280 mil acidentes nas rodovias desde 2014. Site da ABRAMET: [link]. Acesso em: 20 set. 2025.

Agarkar, A. S., Gandhiraj, R., and Panda, M. K. (2023). Driver drowsiness detection and warning using facial features and hand gestures. In 2023 2nd International Conference on Vision Towards Emerging Trends in Communication and Networking Technologies (ViTECoN), pages 1–6.

Basit, M. S., Ahmad, U., Ahmad, J., Ijaz, K., and Ali, S. F. (2022). Driver drowsiness detection with region-of-interest selection based spatio-temporal deep convolutional-lstm. In 2022 16th International Conference on Open Source Systems and Technologies (ICOSST), pages 1–6.

Bekhouche, S. E., Ruichek, Y., and Dornaika, F. (2022). Driver drowsiness detection in video sequences using hybrid selection of deep features. Knowledge-Based Systems, 252:109436.

Chen, J., Fang, Z., Wang, J., Chen, J., and Yin, G. (2022). A multi-view driver drowsiness detection method using transfer learning and population-based sampling strategy. In 2022 IEEE 25th International Conference on Intelligent Transportation Systems (ITSC), pages 3386–3391.

Desai, M. M., Kathad, K., and Modi, N. (2024). Real-time driver drowsiness detection using hybrid cnn-lstm model with facial feature and behavioral analysis. In 2024 4th International Conference on Ubiquitous Computing and Intelligent Information Systems (ICUIS), pages 197–202.

Ebrahim Shaik, M. (2023). A systematic review on detection and prediction of driver drowsiness. Transportation Research Interdisciplinary Perspectives, 21:100864.

Ghoddoosian, R., Galib, M., and Athitsos, V. (2019). A realistic dataset and baseline temporal model for early drowsiness detection. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops.

Lee, C. and An, J. (2023). Lstm-cnn model of drowsiness detection from multiple consciousness states acquired by eeg. Expert Systems with Applications, 213:119032.

National Center for Statistics and Analysis (2017). Drowsy driving 2015 (crash·stats brief statistical summary). Technical Report DOT HS 812 446, National Highway Traffic Safety Administration, Washington, DC.

Serengil, S. and Ozpinar, A. (2024). A benchmark of facial recognition pipelines and co-usability performances of modules. Journal of Information Technologies, 17(2):95–107.

Zhang, F., Bazarevsky, V., Vakunov, A., Tkachenka, A., Sung, G., Chang, C.-L., and Grundmann, M. (2020). Mediapipe hands: On-device real-time hand tracking. arXiv preprint arXiv:2006.10214.