Detecting Driver Drowsiness With Multi-Sensor Data Fusion Combined With Machine Learning
archive: archived pipeline: cataloged verified
Get this paper ↗ (full text — opens at the source; we link to it, we don't host it)
Summary
This research addresses the critical safety issue of drowsy driving, which the National Highway Traffic Safety Administration estimates causes 91,000 police-reported crashes annually, resulting in significant injuries and fatalities. The study is motivated by the underreporting of drowsiness-related accidents and the need for reliable, real-time detection systems. The authors propose a novel Advanced Driver Assistance System (ADAS) that utilizes multi-sensor data fusion—combining visual surveillance and micro-Doppler radar—with deep learning to detect driver fatigue and alert drivers before accidents occur. The system integrates a webcam with night vision capabilities and a 10.5GHz micro-Doppler radar sensor. The webcam captures facial images to monitor eye blinking patterns, yawning incidence, and head position, while the radar detects subtle head movements indicative of nodding off. These inputs are processed by Deep Convolutional Neural Networks (DCNNs) optimized for edge computing. The models were trained on specific datasets: the Closed Eyes in the Wild dataset for blink detection, the IEEE DataPort YAWDD dataset for yawn detection, and custom spectrogram images for radar-based head drop detection. To ensure real-time performance, the TensorFlow models were converted to TensorFlow Lite and quantized for execution on a Raspberry Pi 4 paired with a Google Coral USB Accelerator (Edge TPU). Experimental results demonstrated high accuracy across all detection modules. The blink detection model achieved a validation accuracy of 91.8%, utilizing a threshold of seven consecutive closed-eye frames to distinguish blinks from drowsiness. The yawn detection model reached a validation accuracy of 95.1%. The micro-Doppler radar model, which classifies spectrogram images of head motion, achieved a validation accuracy of 92.11%. The complete prototype was tested in a vehicle environment, where the system continuously assessed driver alertness over 60-second intervals. Upon detecting drowsiness, the system triggered a visual warning on a touchscreen dashboard and activated a vibration motor on the steering wheel. The study concludes that combining visual and radar sensors with machine learning provides a robust solution for drowsy driver detection, achieving over 95% overall accuracy in the working prototype. This approach offers a practical, installable technology for existing vehicles that can operate in various lighting conditions. By providing immediate alerts through visual and tactile feedback, the system aims to reduce the prevalence of drowsy driving accidents, thereby enhancing road safety and potentially saving lives. The research highlights the effectiveness of data fusion in overcoming the limitations of single-sensor systems and demonstrates the viability of edge-based AI for real-time automotive safety applications.
Key finding
The multi-sensor drowsy driver detection system achieved over 95% accuracy in detecting drowsiness by fusing webcam-based facial feature analysis with micro-Doppler radar head movement data.
Methodology
lab_experiment
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 bulk_ingest_rosap on 2026-05-23 (6 acquisition events logged).
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | rosap | — | — | 2 | 2026-05-23 |
| archive | success | — | — | — | 1 | 2026-05-23 |
| 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 | 2026-05-23 |
| promote | success | — | — | — | 1 | 2026-05-23 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 3 | 2026-06-10 |
| tag | success | vector_similarity | — | — | 19 | 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.
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: tool software
- Theoretical Contribution: computational model