Drowsiness monitoring based on driver and driving data fusion

Daza, Iván García; Hernández, Noelia; Bergasa, Luis M.; Parra, I.; Yebes, J. Javier; Gavilan, M.; Quintero, R.; Llorca, David Fernández; Sotelo, Miguel Ángel · 2011 · OpenAlex-citations

DOI: 10.1109/itsc.2011.6082907

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 presents a non-intrusive system for monitoring driver drowsiness by fusing driver-specific visual data with vehicle driving dynamics. The research is motivated by the significant safety risks associated with sleepiness, which increases accident risk four to six times and contributes to 15–20% of vehicle accidents. Existing detection methods are categorized into biomedical signals (intrusive), driving performance metrics (limited by environmental factors and ineffective for micro-sleeps), and computer vision (often requiring complex calibration). This study aims to overcome these limitations by combining the Percentage of Eye Closure (PERCLOS) with driving signals to improve detection accuracy. The experimental design utilized a naturalistic driving simulator featuring a real truck cab and a 180-degree visual field. Data were collected from 10 professional drivers who performed 60-minute driving sessions under two conditions: well-rested and sleep-deprived (four hours of sleep). Ground truth labels for drowsiness were established using the Karolinska Sleepiness Scale (KSS) and subjective expert annotation of video and signal data. The system processed PERCLOS via a real-time stereo vision system and extracted driving signals—lateral position, steering wheel angle, and heading error—from the CAN bus. These signals were analyzed in both time and frequency domains to generate statistical indicators, such as mean and standard deviation, and spectral density power. These features were fed into a two-layered feed-forward artificial neural network (ANN) with 20 hidden neurons to classify the driver’s state. The results demonstrated that individual indicators yielded varying detection rates, with PERCLOS achieving the highest single-indicator rate at 97.61%, followed by the standard deviation of lateral position at 84.10%. Driving signals alone performed poorly, with steering wheel angle reaching only 40.57%. However, fusing indicators significantly improved performance. The combination of PERCLOS and the standard deviation of lateral position achieved the highest detection rate of 98.65%. Other combinations, such as PERCLOS with average lateral position, also yielded high accuracy (97.34%). Conversely, frequency-domain energy indicators and heading error signals did not improve detection rates when fused with other metrics, as they lacked strong correlation with drowsiness patterns. Confusion matrices confirmed high precision, with the best model correctly identifying 100% of awake states and 93.75% of drowsy states. The study concludes that fusing PERCLOS with driving performance metrics, specifically the standard deviation of lateral position, provides a robust and accurate method for drowsiness detection, significantly outperforming individual indicators. This approach addresses the limitations of vision-only systems by incorporating vehicle dynamics, which reflect the driver’s loss of control. The authors note that while the system performs well in a simulator, future work must validate its robustness in real-world driving conditions to ensure practical applicability for intelligent transportation systems.

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.

StageOutcomeToolModelPromptAttemptsCompleted
discover success OpenAlex-citations 1 2026-06-19
archive success semantic_scholar 6 2026-06-26
extract success cached 2 2026-06-26
clean success clean 1 2026-06-19
chunk success chunk 1 2026-06-19
embed success embed Qwen/Qwen3-Embedding-8B 1 2026-06-19
promote success 1 2026-06-19
summarize success llm qwen3.6-27b-prismaquant summ-v5 1 2026-06-26
tag success vector_similarity 6 2026-06-19
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).