Driver State and Behavior Detection Through Smart Wearables
DOI: 10.48550/arXiv.2104.13889
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 limitations of camera-based driver monitoring systems in semi-automated vehicles, which often struggle with low-light conditions, high computational costs, and privacy concerns. The authors propose using passive sensing via off-the-shelf smartwatches to classify driver activities, outside events, and road attributes in naturalistic driving scenarios. This approach aims to provide contextual awareness for shared autonomy systems while respecting user privacy and reducing resource requirements. The researchers utilized data from the HARMONY naturalistic driving study platform, collecting multimodal sensor data from 15 participants using Android smartwatches. The wearable devices captured inertial measurement unit (IMU) data, photoplethysmography (PPG), heart rate, ambient light, noise levels, and GPS coordinates. Ground truth labels for driver activities (e.g., phone use, eating), outside events (e.g., lane changes, intersections), and road types (e.g., city streets, highways) were manually annotated using synchronized in-cabin and exterior video footage. To handle varying sensor frequencies and class imbalances, the data was resampled to 10 Hz, and features were extracted using the tsfresh library. The authors employed Synthetic Minority Oversampling Technique (SMOTE) to balance the dataset and trained Random Forest, Decision Tree, and Extra Tree classifiers using 10-fold cross-validation. The Random Forest classifier demonstrated superior performance across all categories. For driver activity classification, the model achieved an average F1 score of 94.55% on oversampled data, with significant improvements in detecting minority classes like "searching for an item." Outside event classification reached an F1 score of 98.27%, and road type detection achieved 97.86%. Feature importance analysis revealed that accelerometer data was most critical for activity recognition, while PPG and heart rate sensors significantly contributed to detecting road types and outside events, highlighting the value of multimodal sensing. The study also showed that adding sensor modalities incrementally improved classification accuracy. The findings demonstrate that commercially available smartwatches can accurately detect driving context in real-world settings, offering a viable, privacy-preserving alternative to video-based monitoring. This technology can complement vision systems, particularly in low-light conditions or when users prefer higher privacy settings. The authors suggest future work should include expanding participant numbers, integrating additional sensor modalities like skin temperature, and developing real-time edge computing implementations to optimize battery usage and system responsiveness.
Key finding
Smartwatch IMU plus heart-rate features can classify driver activities, exterior events, and road-type context at vision-comparable accuracy (F1 above 0.94) in naturalistic driving — establishing a privacy-preserving alternative to in-cabin cameras.
Methodology
naturalistic
Sample size: N=15
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 | — | — | — | 1 | 2026-05-07 |
| archive | success | canonical_url | — | — | 2 | 2026-06-03 |
| extract | success | cached | — | — | 2 | 2026-06-10 |
| clean | success | — | — | — | 1 | 2026-06-01 |
| chunk | success | — | — | — | 1 | 2026-06-01 |
| embed | success | — | — | — | 1 | 2026-06-02 |
| enrich | success | — | — | — | 1 | 2026-05-07 |
| promote | success | — | — | — | 3 | 2026-06-06 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 2 | 2026-06-10 |
| tag | success | vector_similarity | — | — | 15 | 2026-06-11 |
| verify | partial | — | — | — | 2 | 2026-06-10 |
Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified_with_issues.
Topics
Ranked by relevance to this paper. Hover a topic for its definition.
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: physiological data
- Methodological Resource: validation psychometrics, tool software