无参数局部保持投影及人脸识别  

Parameter-Free Locality Preserving Projections and Face Recognition

作　　者：黄璞[1] 唐振民[1] 

机构地区：[1]南京理工大学计算机科学与工程学院南京210094

出　　处：《模式识别与人工智能》2013年第9期865-871,共7页Pattern Recognition and Artificial Intelligence

基　　金：国家自然科学基金项目(No.90820306、61125305);江苏省普通高校研究生科研创新计划项目(No.CXZZ12_0204)资助

摘　　要：局部保持投影(LPP)通过构造近邻图来保持样本的局部结构,在构图过程中,LPP面临复杂的参数选择问题.为解决此问题,提出无参数局部保持投影(PLPP)算法.首先设计一种无参数的构图方法,能够动态地获取样本的近邻点并配置相应的边权.其次,利用该构图方法,PLPP通过寻求最佳投影矩阵,用于保持样本在低维空间的局部结构.由于PLPP在构图过程中并未设置任何参数且采用余弦距离设置边权,因此PLPP计算更加方便快捷且对离群样本更具鲁棒性.另外,为进一步提升PLPP的识别性能,在PLPP的基础上通过加入样本的类别信息,提出监督的无参数局部保持投影算法(SPLPP).最后,在ORL、FERET及AR人脸库上的实验验证了PLPP与SPLPP的有效性.Locality Preserving Projections （LPP） aims to preserve local structure of the data by constructing a nearest-neighbor graph. However, it is confronted with the difficulty of parameter selection in the process of graph construction. To solve this problem, an algorithm called parameter-free locality preserving projections （PLPP） is proposed. Firstly, a parameter-free graph construction strategy is designed, which can actively determine neighbors of each data point and assign corresponding edge weights. Then, with the proposed graph construction strategy, PLPP seeks an optimal transformation matrix to preserve local structure of the data in the low dimensional space. Since PLPP needs no parameters in graph construction and takes cosine distance as the similarity weight, it is more efficient and robust to outliers than LPP. Moreover, supervised PLPP （SPLPP） is proposed to improve the discriminant ability of PLPP by considering class information of samples. The experimental results on the ORL, FERET and AR face databases validate the effectiveness of PLPP and SPLPP.

关 键 词：人脸识别 特征提取 流形学习 图构建 无参数局部保持投影 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]