Identification of Sleepy Drivers as a Means of Maintaining Roadway Safety

Driver drowsiness is a significant cause of accidents and deaths. The detection of driver tiredness and related indications is a current study topic. Traditional methods are vehicle-, social-, and physiological-based. Many of these approaches are obtrusive, annoying, costly, or require significant data processing. This work attempts to design an economical, real-time, accurate driver sleepiness detection system to address these difficulties. Our cost-effective method uses a camera to capture video and image processing to determine the driver's face in each frame. Face landmarks are determined on the detected face. After that, eye angle, mouth opening, and nose length ratios are computed. A dynamically configurable thresholding method assesses sleepiness based on these data. To improve accuracy, machine learning methods are used offline. Support Vector Machine-based classification has 95.58% sensitivity and 100% specificity. The article employs artificial intelligence and social signals to identify driver drowsiness. Human faces contain a variety of information that may be used to assess tiredness. Facial expressions including eye blinks, head motions, and yawning might indicate exhaustion. However, creating a reliable and effective sleepiness detection system requires strong and accurate algorithms. Prior research has explored several sleepiness detecting approaches. With the rise of deep learning, tiredness detection systems must be reevaluated. Thus, this research examines AI methods including Support Vector Machines, Convolutional Neural Networks, and Hidden Markov Models for sleepiness detection. A full meta-analysis of 25 AI-based sleepiness detection publications is also done. The investigation shows that Support Vector Machines are the most popular technique for sleepiness detection, while Convolutional Neural Networks perform better.