Study Of Face Recognition Methods Based On Ensemble Learning

With the rapid development of computer technology and internet techniques, face recognition is one of the intensive research topic about patten recognition and image processing. For the affect of recognition algorithm and environmental conditions, high-efficiency face recognition rate is the corn issue of relevant research. Since 90s of 20 century, ensemble learning is becoming new research direction of machine learning. Ensemble learning is a kind of multi-learning way which can improve the generalization ability of machine learning. In the application aspect, ensemble learning shows strong power in many fields and obtains certain achievement in the domestic and foreign countries. The application of ensemble learning in face recognition is gradually extending and the inherently distinctive of the system improves distinctly. So, people begin to contact the ensemble learning with face recognition method.The paper researches the face recognition method based on ensemble, for the recognition rate of single classifier is low and to improve the performance of classifiers , it uses ensemble learning concept to fuse the classifiers. Here researches some kinds of classifiers ensemble method. First, researches the application of n tuple classifier on binary image, makes ensemble of the different type n tuple classifier in the experiment, contrasts the recognition results of the method and other face recognition method and the experiment verifies the classification performance of this ensemble method. On the base of n tuple classifier, the idea of bit-plane decomposition further researches face recognition method based on moving window classifier. Analyze the method and uses different ensemble methods to combine classifiers, the experiment verifies the ensemble performance of moving window classifier, contrasts recognition result of the method with other face recognition method and the result shows the performance of the ensemble classifier using sum rule is high. Finally, according to the concept of subspace, the paper researches the ESRS and its face recognition methods, analyze character of different face databases and make experiments on the database. According to the experiments result, it can know the generalization ability of the method is strong and can obtain good result on different databases.