HybridFatigue: A Real-time Driver Drowsiness Detection using Hybrid Features and Transfer Learning
DOI: 10.14569/ijacsa.2020.0110173
archive: archived pipeline: cataloged verified
Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)
Summary
This paper addresses the critical safety issue of driver drowsiness, a leading cause of road accidents, by proposing "HybridFatigue," a real-time detection system. The authors identify significant limitations in existing driver drowsiness detection (DDD) systems, particularly their poor performance during nighttime driving, issues with head position occlusion, and the intrusiveness of physiological sensors. To overcome these challenges, the study aims to develop a robust, non-intrusive system that combines visual and non-visual features using advanced deep learning techniques. The HybridFatigue system employs a hybrid approach integrating visual features extracted via a multi-camera setup and non-visual features obtained from electrocardiogram (ECG) sensors mounted on the steering wheel. The visual component utilizes two cameras—one wide-angle for head tracking and one narrow-angle for eye detection—to mitigate issues with head position and lighting. The system extracts visual features using a pre-trained Convolutional Neural Network (CNN) via transfer learning, while physiological data is captured through ECG sensors. These hybrid features are then classified using a Deep Belief Network (DBN). The model was trained and tested on a combined dataset of 4,250 images sourced from three public datasets: Columbia Gaze Dataset (CAVE-DB), Closed Eyes in the Wild (CEW), and the Multimodality Drowsiness Database (DROZY). The implementation was conducted using Python, OpenCV, Keras, and TensorFlow. Experimental results demonstrate that the HybridFatigue system achieves an average detection accuracy of 94.5% across different subjects and variable environmental conditions. The system classifies driver states into three categories: drowsy, sleepy, and normal. By combining visual PERCLOS measures with ECG heart-beat signals, the system maintains high accuracy even when visual data is compromised, such as during nighttime driving or when drivers wear sunglasses. The use of transfer learning with CNN and DBN architectures allowed for efficient feature extraction and classification, outperforming state-of-the-art single-modality DDD systems. The significance of this work lies in its ability to provide a reliable, real-time solution for driver fatigue detection that addresses the shortcomings of previous vision-only or intrusive sensor-based systems. The hybrid nature of the system ensures robustness against environmental variations and driver behaviors that typically degrade detection accuracy. The findings suggest that integrating multiple data sources with deep learning architectures can significantly enhance the reliability of DDD systems, potentially reducing accident rates and improving road safety. This approach offers a cost-effective and less intrusive alternative for monitoring driver attentiveness in real-world driving scenarios.
Provenance
The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | Crossref | — | — | 1 | 2026-06-19 |
| archive | success | canonical_url | — | — | 1 | 2026-06-26 |
| extract | success | cached | — | — | 2 | 2026-06-26 |
| clean | success | clean | — | — | 1 | 2026-06-20 |
| chunk | success | chunk | — | — | 1 | 2026-06-20 |
| embed | success | embed | Qwen/Qwen3-Embedding-8B | — | 1 | 2026-06-20 |
| enrich | success | openalex | — | — | 1 | 2026-06-20 |
| promote | success | — | — | — | 1 | 2026-06-19 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 1 | 2026-06-26 |
| tag | success | vector_similarity | — | — | 6 | 2026-06-20 |
| verify | partial | — | — | — | 1 | 2026-06-26 |
Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified_with_issues.
Topics
Ranked by relevance to this paper. Hover a topic for its definition.
Information type
What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).
- Empirical Findings: physiological data
- Methodological Resource: tool software