Machine Learning-Based System for Detecting and Tracking Driver Drowsiness

Drivers who aren't paying attention or staying attentive are to blame for many accidents occurring on the world's roads nowadays. The phrase for this condition is "driver drowsiness." This results in a wide variety of undesirable circumstances that endanger human lives and harm them. The key goals of this study are to identify driver sleepiness and to design an appropriate reaction to such findings. There are several ways, many based on the vehicle's movement or the driver's actions. One of the approaches is the physiological method, which assists in diverting the driver's attention away from exhaustion and making him alert. Only some of the solutions entail the use of expensive sensors or a considerable quantity of data. Consequently, this research aims to develop a system that can identify sleepiness in real-time while maintaining an acceptable degree of accuracy. In this method, a camera is used to capture the driver's facial expressions so that they may be examined afterward. A variety of image processing methods are employed to assess each frame for all motion. The landmark locations on the face are used to calculate the Eye Aspect Ratio (EAR), Nose Length Ratio (NLR) & Mouth Opening Ratio(MOR). The derived values are compared to the system's threshold values, and the detection is based on the difference in weight. While this is happening, machine learning algorithms are also being applied offline. Based on the categorization, the system achieved a sensitivity of 98.45% and a specificity of 100% using a Random Forest Algorithm (RFA). This model system is appropriate for use with all sorts of motor vehicles