一种新的模糊稀疏表示人脸识别算法

稀疏表示人脸识别算法的主要思想是:一个未知的测试图像可以近似表示为所有与其隶属同类的训练样本的一个线性组合.然而,人脸之间存在着极大的相似性,同时易受到外部环境的影响,人脸分类的本身存在着一定的不确定性.针对这种不确定性,结合模糊集合理论,提出了一种新的模糊稀疏表示人脸识别算法.首先,引入一个非线性函数描述人脸的相似性程度.然后,基于该相似性度量以及最近邻分类器思想,定义一个自适应的模糊隶属度函数来分配人脸对类的隶属程度.而这一过程恰使得这些隶属度是稀疏化的.最后,将稀疏化的模糊隶属度作为训练样本表示测试样本的权值系数,进而重构测试图像.采用MATLAB在ORL和Yale人脸数据库上进行仿真实验,验证了该算法的有效性和稳定性.

A new fuzzy sparse representation algorithm for face recognition

The main idea of a sparse representation-based face recognition algorithm is that an unknown test sample is approximately represented as a linear combination of all the training samples corresponding to the same class with it. However, the classification of faces may possess some uncertainty, because there is a great similarity among faces and they are easy to be affected by environmental conditions. Based on this uncertainty and fuzzy theory, a new fuzzy sparse representation (FSR) algorithm for face recognition is proposed. Firstly, a new nonlinear function is introduced to represent the similarity among faces. Then, based on the similarity measurement and the nearest neighbors classifier, an adaptive fuzzy membership function is defined, which produces the corresponding membership degree to the class. During this process, the membership degree is sparsity. Finally, the sparse fuzzy membership degree is taken as the weight coefficients of training samples to express test samples to restructure the test image. Utilizing MATLAB, the experiments conducted on the ORL and Yale face databases have demonstrated the effectiveness and robustness of the proposed algorithm.