An Integrated Framework for Driver Drowsiness Detection and Alcohol Intoxication using Machine Learning

Abstract

Road accidents are drastically increasing day by day. Alcohol consumption of the driver and driver drowsiness are major causes for majority of the accidents. Harnessing the powerful technology like machine learning can prevent the two major causes of road accidents; fatigue and alcohol intoxication. Many conventional methods of drowsiness detection are either intrusive or require expensive sensors and data handling. Moreover, there is no inbuilt system to detect drunk driving. Therefore, a low cost, non-intrusive real-time driver’s drowsy detection system integrated with alcohol detection system is developed with relatively higher accuracy and quality. In this system, initially alcohol is sensed using MQ-3 sensor. Subsequently, a web-cam placed on the dashboard of the car is used for face detection (Support Vector Machine and Histogram of Oriented Gradient). Based on the threshold values of four core facial features extracted drowsiness is detected ensuing an alert. Raspberry Pi 3 and Arduino UNO is used to integrate both systems. The accuracy of the system sums up to 86%.