Design and Implementation of Eye Blink Detection System
Dr Venkatesha G1, Tejas Y2, Soumyadeep Das 3, Niranjan N S 4, Mohammed Shoaib5
1HOD, CSE(IoT), East West Institute of Technology, Bengaluru, India
2Student, CSE(AI&ML), East West Institute of Technology, Bengaluru, India
3Student, CSE(AI&ML), East West Institute of Technology, Bengaluru, India
4Student, CSE(AI&ML), East West Institute of Technology, Bengaluru, India
5Student, CSE(AI&ML), East West Institute of Technology, Bengaluru, India
Abstract - Driver drowsiness is a significant contributor to road accidents worldwide, leading to severe injuries, fatalities, and economic losses. This paper presents the design and implementation of a non-intrusive, real-time drowsiness detection system using computer vision. The proposed system utilizes a standard webcam to capture a live video feed of the driver. A modular processing pipeline is employed, beginning with face detection using a Histogram of Oriented Gradients (HOG) with a Linear SVM classifier. Subsequently, Dlib's 68-point facial landmark predictor localizes the eye regions. The core of the methodology involves calculating the Eye Aspect Ratio (EAR), a scalar metric derived from the coordinates of six facial landmarks around each eye, which provides a robust measure of eye openness. By analyzing the temporal sequence of EAR values, the system detects normal blinks and, more critically, prolonged eye closures indicative of microsleep events. A drowsiness score is computed based on blink duration and frequency, and an audible alarm is triggered when a predefined threshold is exceeded. Developed in Python using OpenCV and Dlib, the system is designed to be a cost-effective and accessible solution. Performance evaluation in controlled testing scenarios is designed to achieve high accuracy and a balanced F1-score, demonstrating the system's potential to enhance road safety.
Key Words—Drowsiness Detection, Computer Vision, Eye Aspect Ratio (EAR), Facial Landmarks, Real-Time System, Driver Safety
