Facial Expression Recognition Using Cascaded Random Forest Based On Local Features

Automatic facial expression recognition(FER)is an intriging and challenging topic which has potential applications in natural human-computer interaction because of the unprecedented development of the computer science.Researches in this field have made great progress.However,when it comes to conditional situation,some theoratical algorithms are not pragmatic and versatile,so continuous efforts should be made to further improve the recognition accuracy for practical use.In this paper,an effective method is proposed for FER using a cascaded random forest based on local features.First,the hybrid features of appearance and geometric features are extracted within the salient facial regions sensitive to different facial expressions;second,a cascaded random forest based on the hybrid local features is developed to classify facial expressions in a coarse-to-fine way.We tested our algorithm on both CK+ database and BU3D-FE database,comparing the traditional algoritm's performance on standard database and natural database.Meanwhile,we have proved that the model based on hybrid features have the natural predisposition of higher robust and progressive efficiency towering over the model based on one channel.Extensive experiments also have showed that the proposed method provides better performance comparing to the state of the art on different datasets.