Real-time head pose estimation using weighted random forests

Abstract

In this paper we proposed to real-time head pose estimation based on weighted random forests. In order to make real-time and accurate classification, weighted random forests classifier, was employed. In the training process, we calculate accuracy estimation using preselected out-of-bag data. The accuracy estimation determine the weight vector in each tree, and improve the accuracy of classification when the testing process. Moreover, in order to make robust to illumination variance, binary pattern operators were used for preprocessing. Experiments on public databases show the advantages of this method over other algorithm in terms of accuracy and illumination invariance.