With many researchers’ in-depth study on face recognition,the growth of face recognition technology is very quick.In addition,face recognition technology has the advantages unmatched by some other biometric techniques,such as immediacy,friendliness,and high robustness.The application of face recognition has been gradually extended into everyday life,which substantially improves people’s work efficiency,and also ensures personal or collective information security.However,there are still some issues to be solved in the development of face recognition.For example,the face images are influenced by other factors such as beam,occlusion,expression,posture,skin color and age.As a result,lack of samples resulting in the performance reduction of the algorithms.Many researchers also put forward a lot of solutions,in which the virtual sample generation methods and the sparse representation classification methods are one of the popular methods.Based on the research and analysis of the working principle of virtual sample generation algorithms and sparse representation classification algorithm,this paper puts forward two algorithms for constructing virtual samples to expand the sample set.In addition,three improved sparse representation algorithms are proposed to improve the face recognition rate.First,this paper uses the symmetry of face to design a new virtual sample generation method,and combines the collaborative representation classification algorithm to perform the classification process.The aim of this algorithm is to solve the problem of lack of sample and further improve the effect of face recognition.Secondly,this paper uses the image pixel intensity to design a new description of the image in the form of complex vector,thus producing virtual samples.At the same time,using the collaborative representation classification method with all samples to perform the image classification task.After the deep understanding and analysis of sparse representation classification algorithm,three different improvement algorithms are designed.The purpose of these three methods is to improve face recognition by optimizing the objective function.The proposed algorithms have conducted many comparative experiments on several person face databases or non face databases.In addition,in each experiment,comparing with many most commonly used or latest face recognition algorithms in recent years.Finally,it proves that the algorithms designed in this paper can achieve the purpose of improving the face recognition rate. |