Face Recognition Method Of Identifying Projection And Regression Classification

Face recognition is one of the most important problems in the field of pattern recognition, which has significant impact on theory and application. How to characterize and classify the face is important and difficult for face recognition. Feature extraction based on subspace, one of the ways to characterize face, is based on Statistic Theory and gets satisfying effect. Classifier is another crucial factor for face recognition, because it can influence classification results directly. Feature extraction and classifier are discussed in this dissertation.Collaborative Representation Classifier Steered Discriminative Projection and Wavelet Transform-Projection operator Steered Discriminative Projection are discussed as methods of feature extraction based on subspace. Collaborative Representation Classifier Steered Discriminative Projection aims at finding a subspace in which Collaborative Representation based classification with regularized least square (CRC_RLS) can get the best classification result. Owing to the relation of Wavelet Transform and Projection operator and their good property, Wavelet Transform-Projection operator Steered Discriminative Projection is proposed. Compared with PCA, LDA, LPP, LLE, these two methods get good result on face database.Elastic Net regression based on K-nearest neighbor classifier and Regularization Logistic regression based on K-nearest neighbor classifier belong to classifier based on regression. Li-norm and Lo-norm are add to Least-squares regression and Logistic regression, so called Elastic Net regression model and Regularization Logistic regression model, then these two models can make use of the information of K-nearest neighbor of the sample to classify Compared with other classifiers based on regression, such as SRC. CRC_RLS, the classification performance is improved.