Eye Blinking-Based Drowsiness Detection Using Convolutional Neural Networks

Detecting driver drowsiness plays an important role in improving road safety, as fatigue can lead to slowed reaction time, poor decisions, and reduced awareness. This study presents a real-time drowsiness detection system that uses a Convolutional Neural Network (CNN) to monitor eye states and blinking patterns. The system uses image processing steps such as face and eye detection, converting images to grayscale, and normalizing pixels to prepare the data for training. The CNN model has four convolutional layers, four max-pooling layers, and two fully connected layers, and is trained using the Adam optimizer. The model is trained on a diverse dataset that includes different lighting conditions, head positions, and individual features to make it more adaptable. The performance of the proposed system is tested using fivefold cross-validation, achieving an accuracy of 0.972 on a custom dataset and 0.967 on a dataset collected from Google. The basic approach of the proposed system lies in analyzing the blinking frequency of the eyes and how long the eyes stay closed over several frames. It uses changeable thresholds to adjust the diversified eye blinking rates of different individuals. The proposed system generates real-time audio and visual alerts to warn the car driver when signs of drowsiness are detected. This research highlights the practical value of AI-powered systems in reducing accidents, improving road safety, and protecting drivers by providing early warnings of fatigue.