Driver Drowsiness Detection System Towards Accident Prevention

Abstract

Drowsiness and driver fatigue are the main factors of the road accidents. It is very important to identify driver drowsiness in early stages for minimizing the damage and preventing accidents but it is a very complex and challenging task. It is possible to detect the state of driver fatigue with the development of the technology of computer science. A variety of techniques has been introduced to detect driver drowsiness in the past. In this work, we proposed a novel approach to detect real-time driver fatigue by monitoring behavioral measures on facial expressions, human physiological signals and vehicular parameters. The study addresses the feature extraction methodologies with preprocessing, filtering and normalizing. Facial expressions such as eye features, yawning have been captured and analyzed via computer vision techniques that are based on deep learning algorithms. In this research used built-in sensors of modern wearable devices like smart watches, to extract the signals. Other sensory information such as grip pressure, heart rate, speed of the vehicle, steering wheel behavior has been collected from using specific sensors and simulators. This system designed client-server architecture and that includes several client application modules and main server application. Client modules such as vision application, smart watch app, and grip pressure reader module capture the data from various sensors and send it to the server. These inputs are received at its corresponding server and processed using a drowsiness detection model. That model has been developed with fuzzy rules with modern computer science concepts. The model includes fuzzy rules based on input parameters according to biomedical theories and expert knowledge. The proposed model classify driver drowsiness state into four levels based on input sensory data. Experimental results shows the multi-sensory data and fuzzy model provide valuable contribution for drowsiness detection.