Real-Time Drowsiness Detection For Drivers Using Facial Landmark Analysis And Eye Facet Ratio

Highway safety become a all over world concern to the increasing number of traffic accidents caused by driver fatigue. Drowsiness affects a driver's ability to concentrate, slows reaction time, and impairs decision-making, thereby increasing the likelihood of serious accidents. Fatigue-related incidents are particularly common during extended driving periods, nighttime travel, and monotonous road conditions. As a result, the development of intelligent systems capable of recognizing early signs of driver drowsiness has become an important research topic in the fields of computer vision and intelligent transportation. Such systems can assist drivers by providing timely warnings, thereby reducing the possibility of fatigue-related crashes and improving overall road safety.

Quite a few approaches have planned to detect driver drowsiness, including physiological, vehicle-based, and vision-based techniques. Physiological methods utilize signals such as Electroencephalography (EEG), Electrocardiography (ECG), and heart rate measurements to determine the driver's level of alertness. While these methods generally deliver accurate results, they require wearable devices that may reduce driver comfort during long journeys. Vehicle-based methods analyze steering behavior, lane deviation, and vehicle dynamics; however, their effectiveness depends on road conditions and driving environments. In comparison, vision-based structures have expanded considerable attention because they monitor facial expressions without requiring physical contact, assembly them more applied for real-time applications.

The proposed research introduces a real-time drowsiness detection that analyzes facial features using OpenCV and Dlib's pre-trained 68-point facial landmark model. The system continuously acquires video frames from a webcam, identifies facial landmarks, and calculates Eye Aspect Ratio (EAR) and Mouth Aspect Ratio (MAR) to detect prolonged eye closure and yawning. These capacities are assessed using predefined threshold standards to control the driver's level of alertness. Whenever fatigue-related behavior is recognized, the system immediately produces visual and audio notifications to alert the driver. Owing to its lightweight architecture, low computational requirements, and non-intrusive design, the proposed framework provides an efficient and practical solution for real-time driver monitoring and can serve as a valuable component of intelligent driver assistance systems.

2. LITERATURE REVIEW

Driver drowsiness detection has become an important research topic because of its significant role in reducing fatigue-related road accidents. Many methods have planned to identify driver fatigue by analyzing physiological signals, driving patterns, and facial characteristics. Machine learning-based approaches have demonstrated that combining physiological measurements with vehicle behavior can improve drowsiness prediction accuracy. Though, these means often wearable sensors, making them less applied for continuous real-world applications [1]. Vision-based monitoring systems have emerged as an effective alternative by continuously observing the driver's facial activities using cameras, providing a convenient and non-intrusive solution for fatigue detection [2].

Recent advancements in deep learning for computer vision that further improved the performance of driver monitoring systems. Convolutional Neural Network (CNN)-based approaches have been broadly adopted to recognize eye closure, yawning, and some facial terminologies associated with driver fatigue [3]. The integration of CNN models with facial landmark detection has improved the accuracy of identifying fatigue-related behaviors in real time [4]. Other studies employed conventional ML techniques to analyze eye movements and facial expressions for detecting drowsiness while generating timely warning alerts [5]. Similarly, digital driver monitoring have demonstrated continuous eye monitoring can effectively identify prolonged eye closure and trigger alarm mechanisms to prevent accidents [6].

Recent study has also absorbed on improving the reliability and computational efficiency of fatigue detection systems. Innovative approaches combining facial analysis with additional sensing technologies have improved discovery presentation under different driving conditions [7]. Comprehensive reviews of driver monitoring technologies have highlighted the advantages and limitations of physiological, behavioral, and vision-based approaches while finding chances for future developments [8]. Enhanced DL techniques have further improved facial feature extraction for fatigue recognition [9], whereas hybrid methods combining facial landmark localization deep learning have achieved more robust real-time performance [10]. Motivated by these developments, the proposed work adopts a lightweight vision-based framework that utilizes Dlib's 68-point facial landmark detector together with Eye Aspect Ratio (EAR) and Mouth Aspect Ratio (MAR) analysis to detect driver drowsiness efficiently without requiring wearable sensors or specialized hardware.

3. SYSTEM ARCHITECTURE AND DESIGN

The planned driver sleepiness detection project is developed as a real-time vision-based framework that continuously observes the driver's facial activities using a webcam. The overall architecture is composed of six functional modules: image acquisition, image preprocessing, face localization, facial landmark extraction, drowsiness assessment, and alert generation. Initially, the cam captures live frames of videos, which are processed using OpenCV to improve image quality and prepare them for facial analysis. The processed frames are then supplied to Dlib's frontal face detector, which accurately identifies the driver's face before extracting facial landmark points.

Once the facial landmarks are detected, the coordinates corresponding to the eye and mouth regions are isolated for further analysis. The landmark themes are used to calculate Eye Aspect Ratio (EAR) and Mouth Aspect Ratio (MAR), which represent the grade of eye openness and mouth opening, respectively. The calculated values are continuously monitored and compared with predefined threshold values. If the EAR remains below the specified threshold for a consecutive number of frames or the MAR exceeds its threshold due to yawning, the system recognizes the driver's condition as drowsy. This rule-based evaluation helps distinguish fatigue-related behavior from normal blinking and routine facial movements.

After the driver's fatigue state is confirmed, the alert module immediately activates warning mechanisms to notify the driver. The system displays a visual warning message on the screen while simultaneously producing an audible alarm to regain the driver's attention. Since the planned framework relies only on a standard cam and software-based image processing, it eliminates for wearable sensors or specialized hardware. The addition of OpenCV with Dlib's facial landmark detector enables fast and efficient execution, making the system suitable for real-time driver monitoring and intelligent vehicle safety applications.

Figure 1: System Architecture

Figure 2. Driver is Active

Figure 3. Driver is Drowsy [closed eyes]

Figure 4. Driver is Drowsy [yawning]

CONCLUSION

This research presented a real-time drowsiness  for driver to recognize signs of driver tiredness through facial behavior analysis. The proposed framework employs OpenCV for video processing and Dlib's pre-trained 68-point facial landmark detector to extract facial feature points from live webcam images. By scheming Eye Aspect Ratio (EAR) and Mouth Aspect Ratio (MAR), the system effectively identifies prolonged eye closure and yawning, which are important indicators of drowsiness. Based on predefined threshold values, the system promptly generates visual and audio alerts whenever fatigue-related behavior is detected, helping the driver regain attention and reducing the possibility of accidents.

The experimental evaluation demonstrates that the proposed approach is capable of operating efficiently in real time while requiring only a standard webcam and modest computational resources. Its lightweight architecture and non-intrusive design make it suitable for practical implementation in intelligent system support drivers. Although the system performs reliably under normal operating conditions, its accuracy may be influenced by factors such as poor illumination, facial occlusions, and significant head movements. Future research can enhance the proposed framework by integrating deep learning-based facial analysis, adaptive thresholding, head pose estimation, infrared imaging, and advanced behavioral monitoring techniques to improve robustness and detection performance in diverse driving environments.

REFERENCES

1. Gwak, J., Shino, M., & Hirao, A. (2022). Early Driver Drowsiness Detection Using Machine Learning Techniques from Physiological Signals, Behavioral Parameters, and Driving Conditions.
2. Adochiei, I.-R., Tirbu, O.-I., Adochiei, N.-I., Pericle-Gabriel, M., Larco, C.-M., Mustata, S.-M., & Costin, D. (2023). Drivers' Drowsiness Detection and Warning Systems for Critical Infrastructures.
3. Ahmed, T., Jyoti, O., & Mou, T. H. (2022). Drivers' Drowsiness Detection System Enhancement Using Deep Learning: CNN-based Approach.
4. Ali, N., Hasan, I., Özyer, T., & Alhajj, R. (2023). Driver Drowsiness Detection by Employing CNN and Dlib.
5. Prasath, N., Sreemathy, J., & Vigneshwaran, P. (2021). Driver Drowsiness Detection Using Machine Learning Algorithm.
6. Rusmin, P. H., Osmond, A. B., & Syaichu-Rohman, A. (2021). Design and Implementation of Driver Drowsiness Detection System on Digitalized Driver System.
7. Maximous, G. S., Tarek, S., & Bastawrous, H. A. (2022). Humantenna-Based Driver Drowsiness Detection.
8. Kamran, M. A., Mannan, M. M. N., & Jeong, M. Y. (2019). Drowsiness, Fatigue and Poor Sleep's Causes and Detection: A Comprehensive Study.
9. Ahmed, T., Jyoti, O., & Mou, T. H. (2024). Drivers' Drowsiness Detection System Enhancement Using Deep Learning: CNN-based Approach.
10. Ali, N., Hasan, I., Özyer, T., & Alhajj, R. (2022). Driver Drowsiness Detection by Employing CNN and Dlib.