基于核的广义Foley-Sammon鉴别分析  

Kernel-Based Generalized Foley-Sammon Discriminant Analysis

作　　者：高秀梅[1] 金忠[1] 郭丽[1] 杨静宇[1] 

机构地区：[1]南京理工大学计算机科学系,南京210094

出　　处：《计算机科学》2004年第9期165-168,197,共5页Computer Science

基　　金：国家自然科学基金(编号:600072034)

摘　　要：广义Foley-Sammon线性鉴别分析(GFSDA)是抽取线性特征的有效方法之一。本文提出了基于核的广义Foley-Sammon鉴别分析(KGFSDA)方法,首先建立KGFSDA的模型,然后从理论上给出KGFSDA模型的具体求解方法。分析表明,KGFSDA保留了GFSDA能从整体上考虑经过广义最佳鉴别变换后样本的可分性的优点,更重要的是该方法能够有效地抽取样本的非线性特征,是对GFSDA的进一步拓展。在ORL标准人脸库上的实验结果表明,该方法在识别性能上优于已有的广义Foley-Sammon线性鉴别分析,也比经典的Foley-Sammon线性鉴别分析更有效。It is well known that the Generalized Foley-Sammon linear discriminant analysis(GFSDA)is an effective linear feature extraction method. In this paper, the kernel-based generalized Foley-Sammon diseriminant analysis (KGFSDA)is proposed. Firstly,the KGFSDA model is proposed. Then the solution of KGFSDA model is given and proved. The analysis shows that KGFSDA retains GFSDA's advantage of that,when calculating the generalized optimal discriminant vectors, it considers the separability of the projected set of the samples from a global view point. More importantly,it can extract nonlinear features effectively,so makes a great developing to GFSDA. The experimental results indicate that the proposed method has better classification performances than existing Generalized Foley-Sammon linear discriminant analysis,and much better than the classical Foley-Sammon linear discriminant analysis.

关 键 词：FOLEY-SAMMON 线性鉴别分析 人脸库 广义 抽取 识别性能 模型 法能 保留 标准人 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术] R733.7[自动化与计算机技术—计算机科学与技术]