Exploratory Development of Algorithms for Determining Driver Attention Status

Herbers, Eileen; Miller, Marty; Neurauter, Luke; Walters, Jacob; Glaser, Daniel · 2024 · Crossref

DOI: 10.1177/00187208231198932

archive: archived pipeline: cataloged verified

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

Summary

This study addresses the challenge of accurately detecting driver distraction using real-time data from camera-based Driver Monitoring Systems (DMS) and vehicle kinematics. Motivated by the high prevalence of distraction-related crashes and the need for effective DMS integration in advanced driver assistance systems, the research aims to identify key variables and algorithmic structures that can reliably distinguish between attentive and distracted driving states. The authors note that existing methods often struggle with high false-positive rates and fail to account for the wide variation in driver behavior across different driving environments. The researchers utilized a naturalistic driving dataset comprising video feeds, kinematic variables, and DMS data from 24 drivers over eight days. To establish ground truth, human reviewers subjectively assessed 10-second epochs of driving footage, categorizing attention levels into four tiers: Not Distracted, Slightly Distracted, Moderately Distracted, and Very Distracted. The team developed four buffer-based algorithms (V0–V3) that tracked "AttentionDuration" and "InAttentionDuration" based on glance locations. While the initial algorithm (V0) relied solely on glance data, subsequent versions incorporated vehicle speed to reduce false positives at low speeds, where off-road glances are often necessary for safety. The final algorithm (V3) employed a random parameter search to optimize thresholds for glance weighting and speed factors, aiming to minimize mean-squared error against the ground truth assessments. The results indicated that algorithm V3 achieved the best overall performance, particularly in detecting moderately and very distracted drivers. However, significant challenges remained; algorithms frequently misclassified attentive drivers as distracted, especially during high-speed maneuvers or low-speed scanning. A key finding was the inverse relationship between detection sensitivity and specificity: as algorithms improved at identifying highly distracted drivers, their accuracy in correctly identifying attentive drivers decreased. The study concluded that gaze position and vehicle speed are essential minimum inputs for distraction algorithms, as glance patterns vary significantly by speed. The authors also noted that simple heuristic buffers have limited efficacy and that future improvements likely require sophisticated machine learning models incorporating additional variables such as steering angle and torque to better interpret maneuver-specific glance patterns.

Key finding

The optimal distraction detection algorithm combined ungrouped glance locations and vehicle speed, though it demonstrated a trade-off where improved detection of very distracted drivers corresponded with decreased accuracy in correctly identifying attentive drivers.

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.

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
• attention allocation
• situational awareness
• drowsiness detection algorithms

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
• Theoretical Contribution: conceptual framework