Smart Driver Alert: AI-Driven Perception System for Drowsiness Detection

Abstract

Driver fatigue is a hidden but serious cause of road accidents, often leading to severe injuries or fatalities. This study introduces a computer-based monitoring system that observes the driver’s face in real time to identify early signs of tiredness. The design blends quick, traditional detection methods with modern learning algorithms to analyze expressions and movements, such as frequent blinking, eye closure, and yawning. If the system recognizes a risk, it instantly sends a warning to the driver. Testing with different publicly available driver datasets shows that the approach works quickly and maintains reliable accuracy under a range of conditions.

Introduction

• Driver fatigue is a major cause of road accidents.

• It impairs reaction time, focus, and decision-making.

• The project aims to build a fast and reliable system that detects early signs of drowsiness using a camera and AI, regardless of lighting, face orientation, or driver appearance.

II. Related Work

Three main approaches have been explored:

1. Body Signal Tracking (e.g., heart rate, brain waves): Accurate but needs wearable devices.

2. Facial Behavior Tracking: Detects yawning, blinking, and head movements via camera—non-intrusive and real-time.

3. Driving Pattern Tracking: Analyzes vehicle behavior (lane deviation, sudden swerves).

Recent methods combine facial tracking with deep learning to improve detection accuracy across diverse conditions.

III. Datasets Used

The system was trained and tested using video data capturing both alert and drowsy drivers in various environments:

1. NTHU Drowsy Driver Detection (NTHU-DDD)

   • Infrared and visible-light videos.

   • Includes yawning, eye closure, looking away, with/without glasses.

2. Yawning Detection Dataset (YawDD)

   • Short clips focused on yawning vs. neutral expressions.

   • Varied participants and lighting.

3. UTA Real-Life Drowsiness Dataset (UTA-RLDD)

   • ~30 hours of real driving videos.

   • Shows gradual fatigue levels across diverse drivers and conditions.

These datasets offer realistic and varied scenarios, helping the system generalize well.

IV. Data Description

• Videos include indicators like:

  • Eye openness

  • Blink frequency

  • Head position

  • Mouth activity (e.g., yawns)

• Labels (alert/drowsy) were either pre-provided or manually annotated.

• Data covers day and night driving, various head poses, and lighting conditions.

V. Methodology & Model

The system detects fatigue using a multi-step pipeline:

1. Face & Eye Detection

   • Image-processing locates eyes and face in real time.

2. Fatigue Sign Detection

   • Tracks eye closure duration, blink rate, and yawns.

3. State Classification

   • A deep learning model classifies the driver as alert or drowsy based on features.

4. Advanced Detection

   • Object detection networks may enhance reliability by analyzing the full face.

5. Real-Time Alerts

   • If drowsiness is detected (e.g., eyes closed >15 frames), an alarm is triggered.

The system uses transfer learning for improved accuracy and efficiency.

VI. Key Features

• Real-time detection with minimal delay.

• Works under varied lighting, driver appearances, and head poses.

• Non-intrusive – requires only a camera, no wearable devices.

• Warns driver early, allowing time to stop or rest before a potential accident.

Conclusion

This project shows that it is possible to build a camera-based system that can notice when a driver is starting to lose focus. By mixing quick detection tools with learning models, it can track facial cues linked to tiredness and warn the driver before their safety is at risk. While results are promising, some challenges remain, such as detecting signs when the driver’s face is partially covered or when light levels are very low. Expanding the training data and adding other sensing methods, like steering pattern analysis, could make the system even more dependable in the future.

References

[1] Albadawi, Y., Takruri, M., & Awad, M. (2024). Overview of driver drowsiness detection methods. Sensors, 25(3), 812. [2] Albadawi, Y., Takruri, M., & Awad, M. (2025). Future challenges in AI-based fatigue detection. Applied Sciences, 15(3), 1102. [3] Albadawi, Y., Takruri, M., & Awad, M. (2025). Detecting driver fatigue with transformer models. Scientific Reports, 15(1), 17493. [4] Al-Hammadi, M., & Al-Taani, A. (2020). Predicting driver fatigue from yawning patterns. EPiC Series in Computing, 69, 372–382. [5] Abtahi, S., Omidyeganeh, M., Shirmohammadi, S., & Hariri, B. (2014). Dataset for yawning detection in drivers. ACM MMSys 2014. [6] Ghoddoosian, R., Galib, M., & Athitsos, V. (2019). Real-life driver fatigue dataset and baseline model. CVPR Workshops. [7] NTHU Computer Vision Lab. (n.d.). Dataset for drowsy driver detection. [8] Singh, A. (2022). Face recognition using Haar Cascade in Python. [9] Analytics Vidhya. (2020). Introductory guide to CNN-based image recognition. [10] Global Market Insights. (2023). Report on driver fatigue monitoring market.

Copyright

Copyright © 2025 Amanullah Hidayathullah, Mrs. Sharvani v. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.