Weighted Group Sparse Representation Algorithm For Face Recognition

As a non-contact and long-term effective identity recognition technology, face recognition has been widely used in security monitoring, logging and other occasions. At the beginning of 2015, Yun Ma announced that the corporation of Taobao will adopt the technology of electronic face payment, which reflects that the face recognition technology is closely related to human life, and also shows that the technology has been mature and has deep theoretical framework. The traditional face recognition technology is ubiquitously applied in the controlled environment, including the afomentioned electronic face payment technology, which requires the faces of customers to be front, illumination invariant and with no occlusion. However, in some applications such as security surveillance, the training human faces always have light shadow, attitude change, and object occlusion, whick makes the recognition rate results to be poor. Accordingly, the real environmental application of face recognition technology requires more advanced classifier and feature selection algorithms. This paper analyzes the basic theoretical framework of face recognition technology, focuses on the core content of feture extraction and objective identification, and proposes new algorithms of feature mining and objective identification. The innovative work includes the following aspects:(1) We analyse the disadvantages of the typical sparse representation classifier, including the high computational complexity and the lack of sample distribution prior consideration. We propose a new classifier called weighted group sparse representation classification one(WGSRC1) to classify a query face image by minimizing the weighted mixed-norm(l1,2) regularized reconstruction error with respect to training human images. The group sparse regularization is utilized to incorporate the label information and it promotes sparsity at the group level. According to the similarity between a test image and training images of each group, WGSRC1 gives each group a weight. Our method integrates the locality structure of the data and similarity information between the query sample and distinct classes into l1,2-norm regularization to form a unified formulation. The sparse solution of WGSRC encodes more structure information and discriminative information than other sparse representation methods. Experimental results on four face data sets have shown that the proposed method outperforms state-of-the-art sparse representation based classification methods.(2) We analyse the representation fidelity of traditional sparse-type classifier, which is measured by the l2-norm or l1-norm of the coding residual. Such a sparse coding model assumes that the coding residual follows Gaussian or Laplacian distribution, which may not be effective enough to describe the coding residual in practical FR systems. Meanwhile, the sparsity constraint on the coding coefficients makes the computational cost of SRC very high. In this paper, we propose a new face coding model, namely WGSRC2, which could robustly regresses a given image with regularized regression coefficients. Extensive experiments on representative face databases demonstrate that the WGSRC2 is much more effective and efficient than state-of-the-art sparse representation based methods in dealing with face occlusion, corruption, lighting, and expression changes, etc.(3) This paper presents a dimensionality reduction method that fits WGSRC2 well. WGSRC2 adopts a class reconstruction residual based decision rule, we use it as a criterion to steer the design of a feature extraction method. The method is thus called the WGSRC2 steered discriminative projection(WGSRDP). WGSRDP maximizes the ratio of between-class reconstruction residual to within-class reconstruction residual in the projected space and thus enables WGSRC2 to achieve better performance. WGSRDP provides low-dimensional representation of human faces to make the WGSRC-based face recognition system more efficient.