Face Recognition Algorithm Based On Virtual Samples And Weighted Sparse Representation

Face recognition technology is a very important and active research topic in the field of computer pattern recognition,at the same time,face recognition technology is a very important technology of modern biological information identification.It has a large amounts of applications,containing security,human-computer interface,personnel attendance,access control,social network,tracking criminal suspects,etc.In recent decades,experts and scholars have proposed a variety of face recognition methods,face recognition technology has made breakthrough progress both in theoretical research and application,However,face recognition stability and recognition accuracy are still influenced by factors such as obstructions,light intensity,number of training sample and posture expression changes.Therefore,how to reduce the interference of these factors,and to improve the recognition rate of face recognition is a great difficulty and challenge that the scientific researchers have faced,this is also the focus in research of the paper.In recent years,the sparse representation method has been widely applied in the face recognition field because of its good recognition effect.This paper analyzes the basic principle of sparse representation algorithm and effectively combines the sparse representation algorithm with other methods to improve the accuracy of face recognition.(1)In this paper,the face recognition algorithm of virtual samples is further studied,virtual training samples and virtual test samples were generated by adding random noise to training samples and test samples respectively.This algorithm mainly combines the advantages of virtual samples and sparse representations,the original training sample's and the original test samples are constructed to generate enough virtual training samples and virtual test samples to solve the problem of insufficient samples,proven through a lot of experiments:algorithms based on virtual test samples have better experimental results than algorithms based on virtual training samples.(2)A face recognition algorithm based on virtual test sample weighted sparse representation is proposed,a virtual test sample was constructed by randomly adding noise and geometric symmetry transformation to the original test samples,created new training sample sets using distance information between training samples and test samples.The algorithm can be regarded as a strong classifier consisting of four weak classifiers,that is,the sparse representations of the original test sample and the virtual test sample are regarded as a weak classifier,then the new training sample set is used to construct a strong classifier by using the appropriate function to assign dynamic weight distribution to each test sample,finally,the results are classified into the category of minimum error.A large number of comparison experiments show that this algorithm can effectively improve the accuracy of face recognition,thus reducing the error rate of face database image recognition.