Published Paper Details:

Every year thousands of people in India lose their lives due to traffic accidents. The role of human factor plays a key role in the accidents. In general, the driver fatigue alone accounts for around 25 percent of the road accidents and up to 60 percent of road accidents result in death or serious injury. A main cause of fatigue is sleeplessness or insomnia. So, a drivers' drowsiness state is a major factor in severe road accidents that claims thousands of lives every year. The proposed system algorithm's robust performance in detecting drowsiness through face, the system also incorporates a unique feature to enhance its accuracy. By analyzing the orientation of the driver's head, specifically detecting instances where the head tilts either left or right, the system further refines its assessment of drowsiness. This additional parameter helps capture subtle signs of fatigue, such as the head nodding or leaning characteristic of drowsy driving. The system utilizes a facial land mark algorithm for face detection, minimizing the risk of inaccuracies associated with artificial feature extraction. Leveraging the Dlib toolkit, facial landmarks are identified, and the Eyes Aspect Ratio is introduced as a novel parameter for evaluating drowsiness. A fatigue state classifier, based on trained on Eyes Aspect Ratio data, contributes to the real-time monitoring of the driver. To gauge the gradual onset of fatigue, a variable is introduced, calculated as the number of drowsy frames per unit time. This dynamic assessment adds a temporal dimension to the drowsiness detection process, allowing the system to adapt to changing driver conditions. Simulated driving applications demonstrate the system's ability to quickly detect drowsiness in real-time from 640*480 resolution images at over 20fps, achieving an impressive accuracy of 98.80%.