Beyond Breathalyzers: Towards Pre-Driving Sobriety Testing with a Driver Monitoring Camera
DOI: 10.1109/iv64158.2025.11097459
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 addresses the challenge of detecting alcohol-impaired drivers before they begin driving, aiming to replace or supplement expensive and time-consuming field sobriety tests and breathalyzers. Motivated by the significant societal costs of drunk driving and regulatory pushes for in-vehicle impairment detection, the authors propose automated, pre-driving sobriety tests using sensors already present in modern vehicles: a driver monitoring camera, an instrument cluster, and a steering wheel. The core research question is whether short-duration visuomotor tests can reliably distinguish between sober and intoxicated states using only gaze and eye movement data. The authors designed four candidate tests inspired by the physiological effects of alcohol on eye movements and cognitive processing: Gaze Tracking (simplified horizontal gaze nystagmus), Fixed Gaze (vestibulo-ocular reflex assessment), Silent Reading, and Choice Reaction (divided attention task). These tests were implemented in a driving simulator using a Tobii Pro Spark gaze tracker. An exploratory study was conducted with 50 subjects (20 alcohol-impaired, 30 control). Participants completed two sessions: one sober and one after consuming alcohol to reach a target Blood Alcohol Concentration (BAC) of approximately 0.10%. The study collected 7.2 hours of gaze and vehicle control data. To analyze this data, the authors developed a machine learning pipeline using a time-series foundation model (MOMENT-1) to extract features from paired sober and test-state data, which were then compared by a neural network to predict impairment status. The results indicate that the foundation model outperformed a convolutional neural network trained from scratch, achieving a Balanced Accuracy of 0.57 and an F1 score of 0.54. Performance varied significantly by test type; the Choice Reaction test yielded the best discrimination, while other tests performed near chance levels. The Silent Reading test provided the most beneficial data for learning general-purpose signals due to consistent gaze tracking. The authors noted a wide distribution of performance across individual subjects, suggesting that some individuals exhibit more detectable impairment signs than others. Crucially, the system achieved these results using only 10 seconds of observation per test, demonstrating feasibility for pre-driving deployment. The significance of this work lies in its proof-of-concept for automated, in-cabin sobriety testing that requires no additional hardware beyond standard driver monitoring systems. By leveraging gaze tracking, the approach avoids the domain gaps associated with RGB video or speech recognition and offers better privacy protection. The findings suggest that while current models are not yet robust enough for standalone legal enforcement, they can provide objective feedback to drivers who often underestimate their impairment. The authors conclude that future work should focus on personalizing models to individual drivers and exploring the integration of multiple test modalities to improve detection reliability.
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 author_sweep_intake on 2026-05-28 (2 acquisition events logged).
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | author_sweep | — | — | 3 | 2026-05-28 |
| archive | success | canonical_url | — | — | 7 | 2026-06-09 |
| extract | success | cached | — | — | 2 | 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 | — | — | — | 1 | 2026-05-28 |
| promote | success | — | — | — | 1 | 2026-06-04 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 1 | 2026-06-10 |
| tag | success | vector_similarity | — | — | 15 | 2026-06-11 |
| verify | success | — | — | — | 1 | 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.
- alcohol detection systems
- dui enforcement
- drowsy as impairment
- drowsiness detection algorithms
- alcohol
- dms validation
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).
- Methodological Resource: tool software, validation psychometrics, measurement protocol