Study On Locality Constrained Joint Dynamic Sparse Representation For Face Recognition

In recent years, face recognition technology has become one of the most challenging research topics in pattern recognition and biological recognition fields.It has important value in public security and intelligence surveillance, etc.With the rapid development of face recognition technology, many face recognition methods have been proposed by many researchers. However, many problems such as a large amount of data, high dimensionality, nonlinear are not completely solved in traditional face recognition. Recently, motivated by Compressive Sensing(CS), sparse representation techniques have been drawn wide interest and been successfully applied in signal, image, video processing. Sparse Representation-based Classification(SRC) has received much attention for face recognition. However, face images are often influenced by many factors, such as illumination change, expression, pose and disguise variations and so on. They will decrease the performances of SRC.To address the limitation, we propose a method named Locality Constrained Joint Dynamic Sparse Representation-based Classification(LCJDSRC) in this thesis. First of all, we partition a face image into several smaller sub-images. Then, these sub-images are jointly sparse represented using the proposed locality constrained joint dynamic sparse representation algorithm. At last, the representation results for all sub-images are aggregated to obtain the final recognition result. Our proposed LCJDSRC regards the local matching-based face recognition as a multi-task learning problem. Thus, the latent relationships among the sub-images which are from the same face image are taken into account. Meanwhile, the local structure information of the data is also considered in our algorithm. We evaluate our algorithm by comparing it with other mainstream approaches. Extensive experiments on ORL, Extended Yale B, AR and LFW face databases demonstrate effectiveness of our proposed method.