DOI: 10.17587/prin.17.571-579

Impact of Biometric Features on CNN Performance in Driver Drowsiness Detection

D. A. Skorobogatchenko, D. Sc. (Eng.), Professor, dmitryskor2004@gmail.com, I. M. Grigoriev, Master's Student, grigona4@mail.ru, A. D. Kozyakin Master's Student, kozyakin1024@gmail.com, Volgograd State Technical University, Volgograd, 400005, Russian Federation

Corresponding author: Ivan M. Grigoriev, Master's Student, Volgograd State Technical University, Volgograd, 400005, Russian Federation, E-mail: grigona4@mail.ru

This paper presents a comparative study of two approaches to driver drowsiness assessment from video streams: a classical method based on a convolutional neural network (CNN) and a hybrid method that supplements the neural network classifier with geometric biometric facial features. MobileNetV3-Large is used as the base architecture. Biometric features (EAR, MAR, head orientation angles) are extracted using MediaPipe Face Mesh. An adaptive modality fusion mechanism (Fusion Gate) is proposed, dynamically weighting the contributions of visual and biometric features. The integration of biometric indicators addresses limitations of pure CNN approaches under varying lighting, pose, and occlusions, which are common in real-world driving scenarios. Driver drowsiness is a major contributor to road accidents, with human factors causing over 99 % of incidents and microsleep episodes posing particular risks. Experi­ments on publicly available datasets showed that the hybrid method significantly improves performance, achieving an accuracy of 98.7 % (F1 = 0.987) compared to 96.0 % (F1 = 0.958) for the baseline CNN. The results justify the feasibility of using hybrid architectures in driver state monitoring systems, particularly for real-time embedded applications in advanced driver assistance systems (ADAS). This approach enhances robustness and supports safer intelligent transportation systems by better handling human factors contributing to road accidents, offering a practical solution for mass deployment in vehicles.

Keywords: driver drowsiness detection, convolutional neural networks, MobileNetV3, biometric facial features, Eye Aspect Ratio, Mouth Aspect Ratio, MediaPipe, computer vision; multimodal fusion, driver assistance systems

pp. 571—579

For citation:
Skorobogatchenko D. A., Grigoriev I. M., Kozyakin A. D. Impact of Biometric Features on CNN Performance in Driver Drowsiness Detection, Programmnaya Ingeneria, 2026, vol. 17, no. 10, pp. 571—579. DOI: 10.17587/prin.17.571-579 (in Russian).

References:

1. Kalinkin A. L. Prevalence of excessive daytime sleepiness in the Russian Federation, Neuromuscular Diseases, 2018, vol. 8, no. 4, pp. 43—48. DOI: 10.17650/2222-8721-2018-8-4-43-48 (in Russian).
2. Albadawi Y., Takruri M., Awad M. A review of recent developments in driver drowsiness detection systems, Sensors, 2022, vol. 22, no. 5, article 2069. DOI: 10.3390/s22052069.
3. Ramzan M., Khan H. U., Awan S. M. et al. A survey on state-of-the-art drowsiness detection techniques, IEEE Access, 2019, vol. 7, pp. 61904—61919. DOI: 10.1109/ACCESS.2019.2914373.
4. Howard A., Sandler M., Chu G. et al. Searching for Mobile-NetV3, arXiv:1905.02244. 2019. DOI: 10.48550/arXiv.1905.02244.
5. Chaurasiya A., Sonsale S., Ankur et al. Driver Drowsiness Detection System by Measuring EAR and MAR, International Research, Journal of Engineering and Technology (IRJET), 2021, vol. 8, no. 5, pp. 4255—4260.
6. Herath D., Abeyrathne C., Jayaweera P. Vision-based driver drowsiness monitoring: comparative analysis of YOLOv5—v11 models, arXiv:2509.17498, 2025. DOI: 10.48550/arXiv.2509.17498.
7. Jabbar R., Shinoy M., Kharbeche M. et al. Driver drowsiness detection model using convolutional neural networks techniques for Android application, arXiv:2002.03728, 2020. DOI: 10.48550/arXiv.2002.03728.
8. Hashemi M., Mirrashid A., Beheshti Shirazi A. Driver safety development: real-time driver drowsiness detection system based on convolutional neural network, SN Computer Science, 2020, vol. 1, no. 5, article 289. DOI: 10.1007/s42979-020-00306-9.
9. Salman R. M., Rashid M., Roy R. et al. Driver drowsiness detection using ensemble convolutional neural networks on YawDD, arXiv:2112.10298, 2021. DOI: 10.48550/arXiv.2112.10298.
10. Wolter M., Berrio Perez J. S., Shan M. Blinking beyond EAR: a stable eyelid angle metric for driver drowsiness detection and data augmentation, arXiv:2511.19519, 2025. DOI: 10.48550/arXiv.2511.19519.
11. Cao S., Feng P., Kang W. et al. Optimized driver fatigue detection method using multimodal neural networks, Scientific Reports, 2025, vol. 15, article 12240. DOI: 10.1038/s41598-025-86709-1.
12. Venkateswarlu M., Chirra V. R. R. CNN-ViT: a multi-feature learning based approach for driver drowsiness detection, Array, 2025, vol. 27, article 100425. DOI: 10.1016/j.array.2025.100425.
13. Lugaresi C., Tang J., Nash H. et al. MediaPipe: a framework for building perception pipelines, arXiv:1906.08172, 2019, DOI: 10.48550/arXiv.1906.08172.