Stable Sparse Expression Of Complex Target In The Application Of Face Recognition

Face recognition technology is one of the hottest research topics in the field of image processing, pattern recognition, artificial intelligence,etc,because face recognition has the advantages of non-contact operation, confidentiality, uniqueness, and reliability,makes face recognition occupies a very important position in the field of artificial intelligence.A complete face recognition process includes:image acquisition, image pre-processing,face detection and location,face feature extraction and classifier design, and two most important step are face feature extraction and classifier design. Because of the disadvantage of computational complexity.low robustness, traditional face recognition algorithms are constrained in practical applications.At the same time,due to the advantage of the high sparse, high robustness,the sparse representation algorithm has attracted more and more attention.Sparse expression express maximum the amount of information with minimal semaphore,so we can greatly reduce the amount of data and improve the efficiency of operation on the basis of the information required,and sparse expression has its advantage in classification and identification that other algorithms do not have.this paper researches on face feature extraction and classifier design combining sparse express:(1)face feature extraction:The role of feature extraction is image dimensional reduction and extract images category information.Sparse principal component analysis algorithm proposed in this paper is an improvement of the principal component analysis algorithm,it adds L1penalty term in its regression criterion function,which provides a basis for subsequent classification design.Using sparse principal component analysis for feature extraction can not only largely realize data dimension reduction but also select useful features by the sparse features,and redundant features all drop to zero.(2)classifier design:The role of classifier is to classify the face image feature.This paper presents a combination of sparse classifier(CSRC),the first step of classifier design is the design of over-complete dictionary.this article use unit step dictionaries and2-d Gabor cascade method,The dictionary has the characteristics of a sparse and non-negative,greatly reduces the complexity of the algorithm;then use the basis pursuit denoising algorithm to design classifier.Relative to traditional sparse expression algorithms, this classifiable better,recognition rate higher.The experimental results in ORL, Extended Yale B also prove the effectiveness of the algorithm.