Face Recognition Via 2D Subset Learning Techniques And Sparse Representation Classification

In recent years,the compressed sensing theory has broken through the limitation of the traditional sampling rate,which provides a new way for signal acquisition and signal processing.Compressed sensing has attracted much attention,according to the theory of compressed sensing,some scholars proposed a new face recognition algorithm——Sparse representation classification(SRC).The innovation of SRC algorithm lies in its direct use of the training set of face images,then we can express the test samples with the training samples.Finally,we can extract the classified information from the sparse representation.However,if the SRC algorithm is applied directly to raw face images,the computational complexity is very high.In this paper,we propose an improved algorithm,which is a combination of SRC and two-dimensional feature extraction.Before using the SRC algorithm for face recognition,we apply two-dimensional feature extraction algorithm for feature extraction and dimension reduction.The improved algorithm we proposed in this paper can not only reduce the computational complexity of the algorithm,but also can preserve the structure of image.The main work of this paper includes:firstly,we compare some common compressed sensing reconstruction algorithms,and finally we choose the OMP algorithm as the reconstruction algorithm;secondly,we improve the existing SRC algorithm,and the two-dimensional feature extraction algorithm is added to reduce the computational complexity of the algorithm;Finally,experiments on the AR face database are carried out to verify the feasibility and superiority of the improved algorithm.Through the analysis and comparison of the experimental results,it is proved that the proposed algorithm is feasible,and it is superior to the existing algorithms in the recognition rate and the recognition time.