The Monitoring Method of Driver´s Fatigue Based on Neural Network

This paper presents a new method of detecting driver´s alertness level and fatigue. The system incorporates a function for detecting the positions of the driver´s eyes and mouth from the entire facial image and a function for detecting drowsy driving by monitoring changes both in the open or closed state of the eyes and mouth by using their feature points. Put all the feature points into BP neural networks, adopts area matching arithmetic to recognize driver´s alertness level, fatigue monitor system can recognize that the driver is in decreased alertness state when he/she yawns more times than normal, then the system gives off alarm signal, and awakes the driver; the system can also recognize that the driver is in drowsy state when driver´s eyes blink frequency is lower than standard value, then it gives sharp alarm signal to wake the driver up. It has the characteristic of real time, accuracy. The detection performance of the system was evaluated in laboratory tests and actual driving tests using the alertness index as the criterion. Software was devised for adapting a drowsy driving detection system. Test results confirmed that the effectiveness of an algorithm for detecting drowsy driving on the basis of BP neural network was verified in laboratory tests and driving tests. The recognise accuracy rate reaches to 91.8%. This driver fatigue monitoring system has significant effect to reduce traffic accident.