Real time and non-intrusive driver fatigue monitoring

Zhu, Zhiwei; Ji, Qiang · 2005 · OpenAlex-citations

DOI: 10.1109/itsc.2004.1398979

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 fatigue, a leading cause of fatal traffic accidents, particularly in the trucking industry. The authors argue that existing monitoring systems often rely on single visual cues, which are prone to ambiguity and inaccuracy. To overcome this limitation, the study proposes a real-time, non-intrusive driver fatigue monitoring system that simultaneously extracts and integrates multiple visual cues—eyelid movement, gaze, head movement, and facial expression—along with contextual information to robustly characterize a driver’s vigilance level. The system utilizes remotely located CCD cameras equipped with active infrared illuminators to capture video images of the driver. Computer vision algorithms are employed to extract specific fatigue indicators: Percentage of Eye Closure Over Time (PERCLOS) and Average Eye Closure Speed (AECS) for eyelid movement; Gaze Distribution (GAZEDIS) and Percentage of Saccade Eye Movement (PERSAC) for gaze; Nod Frequency (NodFreq) for head tilts; and Yawn Frequency (YawnFreq) for facial expressions. These visual parameters are fused with contextual data (e.g., sleep history, time of day, physical fitness) using a Bayesian Network model. This probabilistic framework infers the unobservable state of fatigue by combining evidence from multiple sources, thereby reducing uncertainty and resolving ambiguities present in single-cue systems. Validation was conducted in two stages. First, synthetic data demonstrated that the Bayesian Network effectively integrates evidence; while single visual cues rarely exceeded the critical fatigue probability threshold, the simultaneous observation of multiple abnormal visual parameters or contextual factors significantly increased the estimated fatigue probability. Second, real-world experiments with eight human subjects validated the system’s accuracy. The computer vision components showed high precision, with eye tracking achieving a 0.05% false alarm rate and 3D face pose estimation maintaining RMS errors below 2 degrees. In a sleep-deprivation study, subjects performed a psychomotor vigilance task (TOVA) after 25 hours of wakefulness. The results showed a strong correlation between the system’s composite fatigue index and the subjects’ response times, confirming that the index accurately reflects declining alertness. The study concludes that the proposed system provides a robust, reliable, and accurate method for real-time fatigue monitoring. By systematically integrating multiple visual cues and contextual information through a Bayesian model, the system offers superior performance compared to single-cue approaches. This approach represents a significant advancement in non-intrusive, active monitoring technologies for accident prevention, demonstrating feasibility for practical application in enhancing driver safety.

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.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• drowsiness detection algorithms
• gaze based attention detection
• distraction detection algorithms
• drowsiness
• truck driver fatigue
• dms validation

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, measurement protocol