多融合协同表示的人脸识别方法

摘要

In this paper,the virtual samples and inverse representation are integrated into collaborative representation to improve the face recognition accuracy.The relation between collaborative representation and inverse representation is gotten by minimizing the norms of collaborative representation coefficients matrix and inverse representation coefficients matrix simultaneously.So the objective function not only includes the col-laborative representation and inverse representation but also uses the available information of virtual samples. The experiment results on Yale and FERET face databases demonstrate the effective and feasibility of the pro-posed method comparing to SRC,TSPTC and CRC.%将虚拟样本和逆表示融入到协同表示中，通过约束协同系数矩阵和逆表示系数矩阵的范数同时最小化的方式充分体现协同表达与逆表达之间的关联信息，不仅在同一个目标函数体现了协同表示和逆表示，而且也能充分利用虚拟样本的有效信息。在Yale和FERET人脸数据库中的实验表明：提出的多融合协同表示方法在不同的训练样本数下均比SRC、TSPTC和CRC的正确率有大幅的提高，证明了提出方法的有效性和鲁棒性。