Face Recognition Algorithm Based On Weighted Sparse Representation

Face recognition is a very important and active research topic in the field of computer vision,and also an important technology of modern biological information identification.It has a large amounts of applications,containing access control,security,the personnel attendance,social network,human-computer interface,criminal investigation,etc.In the past few decades,thousands of methods have been proposed for face recognition,face recognition has been rapid developed both in theoretical research and real-world applications.However,the recognition rate and stability of face recognition are still influenced by light,posture,facial expression,decorations and the training sample number.Therefore,how to overcome the negative effects of these factors,and to improve the accuracy and stability of face recognition are the huge challenges that the scientific researchers have faced,this is also the focus inresearch of the paper.In recent years,sparse representation based classification has been widely applied in face recognition due to its excellent performance.This paper will briefly analyze the basic principle of sparse representation algorithm,and on this basis,the sparse representation algorithm combining with other methods is proposed to improve the accuracy of face recognition.The main work includes:1)A novel weighted sparse representation method based on virtual test samples for face recognition is proposed in this paper.The presented method includes three steps.Firstly,generat virtual test samples for original test samples,and compute the distance between the test sample and each training sample to build a weighted training set.Secondly,represent the test sample over the weighted training set and compute the residuals of each class.Finally,compute the weight of each test sample,and classify the test sample according to the weighted errors.The method,taking the advantages of sparse representation and the virtual sample,is able to predict the change of the test sample information by generating virtual samples,and to emphasize the diversity of different training samples in representing the test sample.In addition,the weights of the original test sample and of the virtual test sample are automatically and dynamically oatained,instead of manual setting.A large number of experiments show that the proposed method can obtain better classification performance than the other methods.2)In order to further improve the accuracy and the stability of face recognition,a novel joint features classification approach with an external generic set for face recognition is presented in this paper.The presented scheme leverages two local features,Firstly,extract the Gabor feature and the LGBP feature of face image,then Gabor feature based representation with an external generic set and LGBP feature based representation with an external generic set are obtained independently.Finally,a weighted score level fusion scheme is adopted to automatically combine Gabor feature and LGBP feature,and to output the final decision.Three metrics,i.e.,recognition rate,stability and execution time,are investigated in the evaluation of the performance of the presented method.The comprehensive experimental results on AR,FERET and LFW face databases demonstrated that the presented approach can always achieve very satisfactory classification results.