Face Recognition Methods Based On Sparse Representation

With the development of society and the progress of science and technology,people begin to realize the importance of information security.Using the stability of biometric information and individual differences to quickly identify personal identity information has become the preferred method in the field of security,in order to protect the information security of national and citizens.Face recognition system,which has the characteristics of non-contact,friendliness and convenience,has attracted the attention of researchers both domestic and abroad,and has become the most widely used method in biometrics recognition.Face recognition involves multiple fields,such as image processing,pattern recognition,and computer vision,which has high theoretical and practical value.Feature extraction and classification recognition are the key points in the face recognition system.In 2008,Wright introduced sparse description into the face recognition systems,which improve the system's recognition performance.Based on the study of SRC research,two improved algorithms are proposed and verified on the face database.The main contents of this paper are as follows:Firstly,the development process of face recognition technology,the application in real life and the existing problems are briefly expounded.Several typical feature extraction methods in face recognition system are analyzed in detail,including principal component analysis,linear discriminant analysis,local-preserving projection and local binary pattern,and the advantages and disadvantages of related methods are compared.A face recognition method based on block sparse representation based on gradient constraint is proposed.Starting from the sparse representation classification,the principle of sparse representation is mainly described.Blocking the image is beneficial to extracting the features of the image.Gradient information is introduced into the face recognition system,and a face recognition method based on gradient sparse block-based sparse representation is designed.First,the training sample and the test sample are divided into blocks,and then the gradient constraint information is introduced to extract new LBP features to construct a new sample image.Finally,the classification of the test sample is performed using the sparse representation classification method.The proposed method is compared with LR,SRC and DLRR.According to the simulation results,it is found that the proposed method can maintain a high recognition rate without blocking and data corruption.A robust face recognition method based on dictionary decomposition sparse representation is proposed.Constructing a robust dictionary is crucial for face recognition.The dictionary is first decomposed into a specific dictionary,a non-specific dictionary,and a sparse error matrix.To correct the corrupted test data,project the data into the corresponding subspace and learn the projection matrix between the original training data and the specific class dictionary;The eigenface method was used to extract the features of the specific class dictionary and the modified test data;The test samples was classified by regularized SRC.Simulation results showed that the recognition performance improved significantly.