Enhancing Driver Safety through Convolutional Neural Networks for Drowsiness Detection

Driver drowsiness is a foremost reason for road accidents globally. This manuscript presents a driver drowsiness detection system using CNN, utilizing in-car camera video input to assess drowsiness levels. The system includes three main stages: face detection, landmark extraction, and classification. In the first stage, the face recognition algorithm detects the face of the driver from the video feed. In the second stage, landmark extraction is performed to detect and locate the facial landmarks. In the final step, CNN is used to classify facial features. This approach provides a unique and innovative solution that overcomes the limitations of existing methods that rely only on visual data. Using physiological signals such as EEG and EOG provides a more reliable and accurate measure of drowsiness levels, as these signals are directly related to the driver's brain activity and eye movement. Furthermore, the proposed methodology includes a real-time implementation that can provide drivers with alerts and warnings when drowsiness levels reach a dangerous threshold. This feature enhances road safety by preventing accidents resulting from driver drowsiness. The proposed system has been trained and tested on a dataset of drivers with different levels of drowsiness. The results show promising outcomes in detecting drowsiness with an accuracy of 92.25%. The proposed system can be unified into the existing vehicle structures to alert drivers when they become drowsy, reducing the risk of accidents on the road.