基于降维字典的L_2范数人脸识别方法研究  

Research of face recognition based on L_2 norm using dimension-reducing dictionary

作　　者：张青苗[1] 殷爱菡[1] 姜辉明[1] 

机构地区：[1]华东交通大学信息工程学院,江西南昌330013

出　　处：《计算机工程与设计》2013年第10期3625-3628,共4页Computer Engineering and Design

基　　金：国家自然科学基金项目(61163040)

摘　　要：基于稀疏表示的人脸识别算法具有良好的识别效果,然而工作机制并不明确。因此,在研究SRC的方法和L1范数作用基础上,提出了一种结合降维字典和L2范数求解的分类方法。利用PCA技术对原始字典进行改进,有效降低原始字典的维数,降低求解系数向量的复杂性;利用L2范数快速计算的特点,加快人脸识别速度,提高系统的鲁棒性和识别率。在Yale B和ORL数据库上的实验结果表明了该方法的有效性。The mechanism of face recognition based on sparse representation is not explicit,even though it has favorable effect.Therefore,a classification scheme is presented,which adopts the combination of dimension-reducing dictionary and L2 norm on the basis of SRS and L1 norm.This scheme optimizes the original dictionary by introducing PCA technique,thus,the dimension of the original dictionary and the computational complexity of coefficient are reduced; meanwhile,the recognition speed and the recognition rate are improved as well as the system robustness,because of the rapid calculation of L2 norm.The validity of this scheme is proved by the experimental results in Yale B and ORL database.

关 键 词：人脸识别 稀疏表示 降维字典 主成分分析法 L2范数 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]