EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

A Comparative Evaluation of Driver Drowsiness Detection Techniques

Wedad M. Albalawi, Sahar Sabbeh, Nuha Zamzami

Engineering, Technology & Applied Science Research · 06.06.2026 · 16 (3) · 35537-35547

DOI: 10.48084/etasr.17766 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Engineering, Technology & Applied Science Research, online erschienen am 06.06.2026. Verfasst von Wedad M. Albalawi und 2 weiteren Autor:innen. Die Arbeit nennt 43 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Drowsy driving is a leading cause of fatal road accidents, underscoring the need for accurate, efficient automated detection systems. Despite significant breakthroughs in deep learning approaches, there is still no clear consensus in the scientific community on whether end-to-end deep learning models or hybrid approaches, which decouple feature extraction from classification, offer more efficient performance. The present study aims to fill this gap by providing a comparative analysis of multiple algorithmic configurations, considering combinations of deep pre-trained feature extractors, such as Visual Geometry Group-16 (VGG16), Mobile Networks (MobileNet), and Extreme Inception (Xception), and traditional feature extractors, such as Histogram of Oriented Gradients (HOG), Local Binary Patterns (LBP), Scale-Invariant Feature Transform (SIFT), and Gabor filters, with a wide variety of classifiers, such as machine learning-based classifiers like Support Vector Machines (SVMs), Random Forest (RF), and Naïve Bayes (NB), and deep learning-based classifiers like LeNet, VGG16, AlexNet, and Convolutional Neural Networks (CNNs). All experiments were conducted on three benchmark datasets: National Tsing Hua University-Driver Drowsiness Detection (NTHU-DDD), Yawning Detection Dataset (YAWDD), and Yawn and Eye Close. The experimental results reveal two main findings. First, hybrid architectures consistently outperform fully end-to-end deep models. For example, the VGG16 network combined with a LeNet classifier achieved 99.32% accuracy on the controlled YAWDD dataset and 98.69% on the real-world NTHU-DDD dataset, whereas MobileNet combined with SVM achieved 99.31% and 98.83%, respectively. Second, incorporating multimodal features (eye and mouth regions) significantly improves performance compared to using either modality alone, highlighting the importance of feature complementarity in drowsiness detection.

Zitieren

APA

Albalawi, W. M., Sabbeh, S., & Zamzami, N. (2026). A Comparative Evaluation of Driver Drowsiness Detection Techniques. Engineering, Technology & Applied Science Research, 16(3), 35537-35547. https://doi.org/10.48084/etasr.17766

Vancouver

Albalawi WM, Sabbeh S, Zamzami N. A Comparative Evaluation of Driver Drowsiness Detection Techniques. Engineering, Technology & Applied Science Research. 2026;16(3):35537-35547. doi:10.48084/etasr.17766

BibTeX
@article{Albalawi2026,
  author = {Albalawi, Wedad M. and Sabbeh, Sahar and Zamzami, Nuha},
  title = {A Comparative Evaluation of Driver Drowsiness Detection Techniques},
  journal = {Engineering, Technology & Applied Science Research},
  year = {2026},
  volume = {16},
  number = {3},
  pages = {35537--35547},
  doi = {10.48084/etasr.17766}
}
Kostenlos starten →