Improving Methods to Measure Attentiveness through Driver Monitoring

Miller, Marty; Herbers, Eileen; Walters, Jacob; Neurauter, Luke · 2022 · ROSA P / Safety through Disruption (Safe-D) University Transportation Center (UTC)

archive: archived pipeline: cataloged verified

Get this paper ↗ (full text — opens at the source; we link to it, we don't host it)

Summary

This research addresses the critical safety issue of driver inattention, which contributes significantly to motor vehicle crashes and economic costs. The study was motivated by the increasing prevalence of Level 2 advanced driver assistance systems (ADAS), which can exacerbate eyes-off-road time, and upcoming regulatory requirements for driver monitoring systems (DMS). The primary goal was to develop and validate algorithms capable of accurately distinguishing between attentive and inattentive driver states using eye-tracking data and vehicle metrics, thereby improving driver trust and system efficacy. The researchers utilized naturalistic driving datasets from General Motors, containing kinematic variables and DMS data from infrared cameras. To establish ground truth, they created a benchmark dataset by manually reviewing 10-second driving epochs, categorizing drivers into four levels of distraction: Not Distracted, Slightly Distracted, Moderately Distracted, and Very Distracted. The study compared frame-by-frame video reduction against DMS outputs, identifying that DMS accuracy was compromised by driver postures, steering wheel obstruction, and eyewear. Algorithm development focused on two approaches: buffer-based algorithms and machine learning models. The buffer-based method used weighted glance locations and speed thresholds to calculate attention status, refined through a random parameter search. Machine learning approaches included neural networks incorporating glance location, speed, steering torque, brake, and throttle data, as well as ordinal logistic regression models. The results indicated that vehicle speed was a crucial factor for understanding driving context and improving algorithm accuracy, as initial models performed poorly at low speeds. The buffer-based algorithms demonstrated fewer false positives compared to deep learning counterparts. While neural networks were more effective at identifying moderately and very distracted drivers, they frequently misclassified attentive drivers as highly distracted. The study confirmed that driver glance locations were vital for determining attentiveness, but also highlighted that DMS hardware limitations, such as obscured camera views, significantly impact algorithm reliability. The significance of this work lies in its contribution to the development of robust, accurate inattention detection methods necessary for the safe deployment of ADAS and compliance with emerging DMS regulations. By demonstrating that combining glance behavior with vehicle context like speed improves detection, the findings provide a framework for designing algorithms that minimize false alarms. This is essential for maintaining driver trust in monitoring systems, ensuring that warnings are perceived as valid rather than ignored, ultimately enhancing roadway safety for all users.

Key finding

Buffer-based algorithms incorporating vehicle speed and glance location weighting produced fewer false positives than deep learning models, though neural networks were more effective at identifying severe distraction.

Methodology

naturalistic

Sample size: 24

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. Discovered via bulk_ingest_rosap on 2026-05-23 (6 acquisition events logged).

Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified.

Topics

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

• gaze based attention detection
• distraction detection algorithms
• visual
• dms validation
• drowsiness detection algorithms
• attention allocation

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: behavioral performance data
• Methodological Resource: tool software, measurement protocol