Recognition of Manual Driving Distraction Through Deep-Learning and Wearable Sensing
DOI: 10.17077/drivingassessment.1670
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 critical safety issue of driving distraction, specifically focusing on the recognition of manual distractions where drivers remove their hands from the steering wheel. Motivated by the high incidence of distraction-related accidents and the growing need for automated driving systems to accurately assess driver attention for safe control transfers, the researchers aimed to develop a novel framework using deep learning and wearable sensors. The study posits that manual distraction, often a precursor to other distraction types, can be effectively monitored through the unique temporal and spatial kinematic patterns of driver movement. The experimental design involved 20 participants driving in a fixed-base simulator across city street and highway scenarios. Participants performed five specific manual distraction tasks—cell phone talking, texting, drinking water, using a navigation panel, and placing a marker in a cup holder—prompted by verbal instructions. Motion data from the right wrist and trunk were captured using wearable inertial measurement units (IMUs). The researchers extracted kinematic segments corresponding to the initiation phase of each task (from hand removal to initial contact with the target) to ensure consistency across participants. A total of 2,342 normalized data segments, including a baseline "DrivingOnly" category, were processed. A modified Convolutional Neural Network (CNN), based on the AlexNet architecture, was trained on data from 16 participants (72% of the dataset) and evaluated on data from the remaining four participants (28%). The model was adapted to handle 7-channel time-series input representing linear position and angular orientation. The results demonstrated that the CNN could accurately classify the type of manual distraction based on right wrist motion. The model achieved an overall accuracy of 87.0% and an F1-score of 0.87 on the testing dataset. Specific recall rates varied by task, with "Drink" (94%) and "Marker" (93%) showing higher recognition rates compared to "Phone" (72%). The confusion matrix indicated that while the model performed well, some misclassifications occurred, particularly between similar hand movements. The study confirms that deep learning techniques can effectively distinguish between different manual distraction behaviors using wearable sensor data. The significance of this work lies in its demonstration of a robust, non-intrusive method for real-time monitoring of driver inattention. By utilizing wearable IMUs, the approach avoids the need for vehicle retrofitting and allows for tracking across different vehicles. The findings suggest strong potential for integrating such systems into advanced driver assistance systems or automated vehicle handover protocols. However, the authors note limitations, including the small sample size and the use of fixed time windows, suggesting that future work should focus on larger datasets, dynamic window lengths, and potentially simpler or more advanced neural network structures to enhance robustness and generalizability.
Key finding
A convolutional neural network trained on wearable inertial measurement unit data achieved 87.0% accuracy in recognizing manual driving distractions.
Methodology
simulator
Sample size: 20
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.
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | author_sweep | — | — | 2 | 2026-05-28 |
| archive | success | canonical_url | — | — | 1 | 2026-06-04 |
| 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 | — | — | — | 1 | 2026-05-28 |
| promote | success | — | — | — | 1 | 2026-06-04 |
| 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).
- Methodological Resource: tool software
- Theoretical Contribution: conceptual framework, computational model