What Can Be Predicted from Six Seconds of Driver Glances?
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 the predictive power of driver gaze patterns, specifically asking what information about driver state, behavior, and the driving environment can be inferred from just six seconds of macro-glance data. Motivated by the need for robust, real-time human-machine interfaces in increasingly automated vehicles, the authors explore whether coarse gaze regions—defined by context rather than precise coordinates—can serve as a reliable sensor for inferring variables beyond simple visual attention. The work aims to demonstrate that short sequences of gaze data can predict complex states such as fatigue, distraction, and environmental conditions, thereby facilitating better vehicle assistance systems. The researchers utilized data from the 100-Car Naturalistic Driving Study, a large-scale dataset comprising approximately 2,000,000 vehicle miles and 43,000 hours of driving from 241 drivers. They analyzed 4,816 randomly selected six-second epochs of baseline driving, which were manually annotated for macro-glances across eight regions (e.g., forward roadway, mirrors, instrument cluster). To model the temporal dynamics of these glances, the authors discretized the sequences into 25 state samples spaced 250 milliseconds apart. They employed Hidden Markov Models (HMMs) for binary classification tasks, training separate models for each class to compute log-likelihoods. This approach was chosen over traditional classifiers like Random Forests or Support Vector Machines, which performed worse, and over explicit-duration models like Hidden Semi-Markov Models, which failed due to the uniform length of the training sequences. The study evaluated 27 binary classification problems covering driving environment variables (e.g., intersection proximity, lighting, traffic density), driver demographics (age, gender), and driver state/behavior (e.g., fatigue, talking, failure to signal). The results indicate that six-second sequences of macro-glances contain significant discriminative signal for predicting several critical variables. The HMM-based classifier successfully predicted driver behaviors such as radio tuning, fatigue, failure to signal, and talking. It also effectively inferred environmental factors, including proximity to intersections, lighting conditions, and traffic density. The analysis revealed significant aggregate differences in macro-glance transition probabilities between different classes, demonstrating that gaze dynamics vary systematically based on the driver’s state and surroundings. For instance, the transition matrices for "not distracted" versus "adjusting radio" showed distinct patterns that the models could leverage for accurate classification. The significance of this work lies in its demonstration that coarse, context-based gaze data can robustly predict a wide range of driver and environmental states in real-world conditions. By proving that macro-glances are sufficient for these predictions, the study supports the development of practical, real-time driver assistance systems that do not require the high-precision, error-prone detection of micro-glances. This approach offers a viable method for understanding the "human" in human-to-vehicle interactions, enabling vehicles to adapt to driver states and environmental contexts more effectively. The findings pave the way for gaze-based interfaces that are resilient to the messy conditions of naturalistic driving, enhancing safety and usability in automated vehicles.
Key finding
Six-second sequences of coarse driver macro-glances can be used to accurately predict specific driver behaviors, states, and environmental variables through Hidden Markov Model classification.
Methodology
naturalistic
Sample size: 241
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 openalex_abstract on 2026-05-08 (2 acquisition events logged).
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | — | — | — | 1 | 2026-05-07 |
| archive | success | canonical_url | — | — | 9 | 2026-06-06 |
| extract | success | cached | — | — | 3 | 2026-06-10 |
| clean | success | clean | — | — | 1 | 2026-06-04 |
| chunk | success | chunk | — | — | 1 | 2026-06-04 |
| embed | success | embed | Qwen/Qwen3-Embedding-8B | — | 1 | 2026-06-04 |
| enrich | success | openalex | — | — | 2 | 2026-05-08 |
| promote | success | — | — | — | 1 | 2026-05-07 |
| 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
- temporal
- visual
- anticipation
- eye movements scanning
- distraction 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, measurement protocol