|Table of Contents|

Theory and Algorithm of Uncorrelated Optimal Discriminant Analysis(PDF)

《南京理工大学学报》（自然科学版）[ISSN:1005-9830/CN:32-1397/N]

Issue:
2002年02期
Page:
179-182
Research Field:
Publishing date:
2002-04-30

Info

Title:
Theory and Algorithm of Uncorrelated Optimal Discriminant Analysis
Author(s):
YangJian YangJingyu LiuNingzhong
Department of Computer Science and Technology,NUST,Nanjing 210094
Keywords:
statist ical uncorrelation;  linear analysis;  feature ext raction;  numerical recogn-i t ion;  discriminant vectors
PACS:
O212
DOI:
-
Abstract:
Some propert ies of the eigenvectors of the generalized eigen-equat ion corresponding to Fisher criterion are analyzed. On the basis of these propert ies, the uncorrelated optimal discriminant vectors are theoret ically proved to be the eig envectors corresponding to the first d largest eigenvalues of the generalized eigen-equat ion. It is st ressed that uncorrelated optimal discriminant analysis ( U ODA) is further development of the classical Fisher linear discriminant analysis in theory . T he results of ex periment on Concordia University CENPARMI handw ritten numerical database indicate the proposed alg orithm is more eff icient than the previous ones, and, U ODA is more powerful than the t radit ional Foley-Sammon discriminant analysis.

References:

1 Tian Q. Image classification by the Foley-Sammon transform. Optical Engineer ing , 1986, 25( 7) : 834~ 839
2  Jin Zhong , Yang J Y, Hu Z S, Lou Z. Face recognition based on uncorr elated discriminant transformation.Pat tern Recog nit ion, 2001, 33 ( 7) : 1 405~ 1 416
3  杨健, 杨静宇, 金忠. 最优鉴别特征的抽取及图像识别. 计算机研究与发展, 2001, 48 ( 11) :1 331~ 1 336
4  胡钟山, 娄震, 杨静宇, 刘克, 孙靖夷. 基于多分类器组合的手写体数字识别. 计算机学报,1999, 22 ( 4) : 369~ 374

Memo

Memo:
-
Last Update: 2002-04-30