一种新的模糊稀疏表示人脸识别算法

发布时间：2018-05-29 04:39

本文选题：人脸识别 + 模式识别　；参考：《大连理工大学学报》2017年02期

【摘要】：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.
[Abstract]:The main idea of the sparse representation face recognition algorithm is that an unknown test image can be approximately represented as a linear combination of all training samples belonging to the same class. However, there is a great similarity between faces, and it is easy to be affected by the external environment. A new fuzzy sparse representation face recognition algorithm is proposed based on fuzzy set theory. Firstly, a nonlinear function is introduced to describe the similarity of faces. Then, based on the similarity measure and the idea of nearest neighbor classifier, an adaptive fuzzy membership function is defined to assign the membership of face to class. This process makes these degrees of membership sparse. Finally, the sparse fuzzy membership degree is used as the training sample to represent the weight coefficient of the test sample, and then the test image is reconstructed. Simulation experiments on ORL and Yale face database with MATLAB are carried out to verify the effectiveness and stability of the algorithm.
【作者单位】：大连理工大学电子信息与电气工程学部控制科学与工程学院;黑龙江科技大学计算机与信息工程学院;
【基金】：国家自然科学基金资助项目(61175041)
【分类号】：TP391.41

【相似文献】