Age Estimation Based On Random Forest

Age is an important factor in face recognition, the first motivation of studying ageestimation is to improve the accuracy of face recognition. With the deepening of age estimationresearch, people find age estimation has broad application prospects and very importanttheoretical value. Therefore, age estimation based on the facial image becomes one hot topic inthe field of pattern recognition and computer vision. So far, many domestic and foreign expertsand scholars have put forward a lot of age estimation algorithms based on the facial image. But,how to further improve the robustness and the accuracy of age estimation, and reduce time andspace complexity of the algorithm, is still a problem that urgently need to be solved. RandomForest method has been widely applied in many fields, such as economics, b iology medicine, andhas shown good performances in these areas. In this paper Random Forest is applied to ageestimation problem first time, a new algorithm for age estimation based on biologically inspiredfeature and random forest is presented. The main contributions are as follows:1. An algorithm for age estimation based on biological inspired feature and Random Forestis presented. First, the biological inspired features of face image is extracted, then, the OLPPmethod is used to reduce the dimensionality of the feature extracted above, finally, random forestregression is used to estimate the age. The algorithm is implemented with Matlab, and is testedon the famous FG-NET face image database.2. The biologically inspired feature is compared with two kinds of widely used features inthe field of computer vision, LBP feature and the HOG feature. The algorithm for age estimationbased on LBP and random forest, and algorithm base on HOG and random forest areimplemented. The algorithms are tested on FG-NET database. The experimental results show thesuperiority of biologically inspired features to LBP feature and HOG feature in age estimation.3. Random forest method is compared with SVM method. Three algorithms for ageestimation based on SVM are implemented in which biologically inspired features, LBP featuresand HOG features are used respectively. The experiments are done on FG-NET database, and theresults show the superiority of random forest method in age estimation.