Investigating the correspondence between driver head position and glance location
DOI: 10.7717/peerj-cs.146
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 investigates whether driver head pose can serve as a reliable surrogate for eye gaze location, addressing the limitations of eye-tracking technology in production systems. The research aims to determine if head rotation data can predict glance locations during on-road driving and to assess the impact of individual differences on this correspondence. The authors utilized secondary data from the Virginia Tech Transportation Institute’s naturalistic driving study, focusing on 22 participants performing dynamic driving tasks. Data collection involved video recording of drivers’ faces to estimate head rotation (X, Y, and Z axes) via facial landmark annotation, alongside manual coding of glance locations into 16 predefined categories. The methodology employed Principal Component Analysis (PCA) to identify salient patterns in head rotation data and tested four machine learning classifiers—k-Nearest Neighbor, Random Forest, Multilayer Perceptron, and Hidden Markov Models (HMM)—to predict glance locations. To address class imbalance, where forward roadway glances vastly outnumbered others, the researchers used subsampling to create balanced datasets. Model performance was evaluated using accuracy, F1-score, and Cohen’s Kappa statistic across 50 Monte-Carlo iterations. Results indicated that head rotation data is a robust predictor of glance location, particularly when the visual angle between glance targets increases. For the binary classification of forward roadway versus center stack glances, the HMM achieved the highest accuracy of 83% on balanced data. Classification accuracy improved significantly with larger visual angles; for instance, distinguishing forward glances from right mirror glances yielded up to 90% accuracy using Random Forest. PCA revealed that X (vertical) and Y (horizontal) rotations were the primary contributors to classification, with Y rotation being most significant for lateral glances. The study also identified substantial individual differences in head-glance correspondence, noting that drivers who exhibited wider ranges of horizontal head rotation tended to have greater mean differences in rotation angles between forward and center stack glances. The findings suggest that inexpensive head pose tracking can effectively estimate driver gaze, particularly for high-eccentricity glances that are critical for distraction detection. While individual variability exists, classifier models based on head rotation remain robust enough to serve as reasonable estimators. This supports the integration of head pose tracking into vehicle safety systems to mitigate driver inattention, offering a practical alternative to complex eye-tracking hardware.
Key finding
Machine learning classifiers, particularly Hidden Markov Models, can accurately estimate driver glance locations from head rotation data, with performance increasing as the visual angle between glance targets widens.
Methodology
naturalistic
Sample size: 22
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 | 2026-06-05 |
| archive | success | openalex | — | — | 5 | 2026-06-06 |
| extract | success | cached | — | — | 3 | 2026-06-10 |
| clean | success | clean | — | — | 1 | 2026-06-05 |
| chunk | success | chunk | — | — | 1 | 2026-06-05 |
| embed | success | embed | Qwen/Qwen3-Embedding-8B | — | 1 | 2026-06-05 |
| promote | success | — | — | — | 1 | 2026-06-05 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 2 | 2026-06-10 |
| tag | success | vector_similarity | — | — | 15 | 2026-06-11 |
| verify | success | — | — | — | 2 | 2026-06-10 |
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
- visual
- distraction detection algorithms
- temporal
- attention allocation
- eye movements scanning
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: measurement protocol, tool software