基于镜像奇异值分解的单样本人脸识别  

Single-sample face recognition based on mirrows and singular value decomposition

作　　者：王梅花[1] 班锦[2] 

机构地区：[1]西南交通大学信号与系统四川省重点实验室,四川成都610031 [2]新乡医学院三全学院,河南新乡453700

出　　处：《微型机与应用》2010年第16期37-39,43,共4页Microcomputer & Its Applications

基　　金：中国国家自然科学基金(60971104);中国高等教育博士项目基金(20090184110008);四川省青年科学技术基金(09ZQ026-091)

摘　　要：目前有许多正面人脸的识别方法,当有充分数量的训练样本时,能取得较好的识别效果,然而当处理单样本人脸识别问题时,效果则明显下降。针对这种情况,提出了基于镜像奇异值分解的单样本人脸识别方法,通过采用镜像的方法增加训练样本信息。实验表明,在对人脸图像进行识别时取得了较好的效果,并且在一定程度上克服了单样本条件下姿态变化对识别效果的影响。Nowdays, there are many methods to deal with face recognition, when there are a sufficient number of training samples these methods can achieve good recognition results, but to deal with a single sample face reeognition problem, the effeets decreased rapidly. To overcome this problem, a new method which is called image-based singular value decomposition is presented, through the use of mirrors to increase the imformation of training samples. Experiments show that the proposed approach is better than the others, and to some extent, it can also overcome the effect of changes in attitude.

关 键 词：训练样本 镜像 奇异值分解 姿态 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]