A monitoring method of driver mouth behavior based on machine vision

When we use the computer vision to inspect the driver´s driving behavior, the identifying of the mouth state is one of the key technologies. In fact, when a driver drives in a normal, talking or dozing state, his/her mouth opening degree will be quite different. According to this fact, this paper uses the Fisher classifier to extract the mouth shape and position, then uses the mouth region´s geometry character as the feature value, and put all of these features together to make up an eigenvector as the input of a three-level Bp network, then we get the output among three different spirit states. The experiment results show that this new method can inspect the driver´s mouth region state accurately and quickly.