[1]束婷婷,甘岚,杨静宇.求解统计不相关的最佳鉴别矢量的统一算法[J].南京理工大学学报(自然科学版),2002,(03):290-294.
 ShuTingting GanLan YangJingyu.A Unified Algorithm for the Computation of Statistically Uncorrelated Optimal Discriminant Vectors[J].Journal of Nanjing University of Science and Technology,2002,(03):290-294.
点击复制

求解统计不相关的最佳鉴别矢量的统一算法()
分享到:

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

卷:
期数:
2002年03期
页码:
290-294
栏目:
出版日期:
2002-06-30

文章信息/Info

Title:
A Unified Algorithm for the Computation of Statistically Uncorrelated Optimal Discriminant Vectors
作者:
束婷婷甘岚杨静宇
南京理工大学计算机科学与技术系, 南京210094
 华东交通大学计算机学院, 南昌330013
Author(s):
ShuTingting GanLan ① YangJingyu
Department of Computer Science and Technology,NUST,Nanjing 210094)
关键词:
最佳鉴别矢量 统计不相关 模式识别 人脸识别
Keywords:
optimal discriminant vectors stat ist ical uncorrelat ion pat tern recognit ion face recognit ion
分类号:
TP391.41
摘要:
Fisher最佳鉴别准则是高维模式分析中的有效方法 ,其关键是求解最佳鉴别矢量。统计不相关的最佳鉴别矢量保证模式矢量投影后得到的特征是统计不相关的 ,已有的计算统计不相关的最佳鉴别矢量算法不能计算小样本的情形 (类内散布矩阵是奇异的 ) ,针对这种情形 ,该文给出了一种对大小样本都能精确计算统计不相关最佳鉴别矢量的统一算法。在大样本情形下 ,该方法得到的结果与已有的方法相同。为验证算法的有效性 ,将其用于人脸识别实验 ,该方法比已有的方法能得到更高的识别率
Abstract:
Fisher opt imal discriminant criterion is an ef ficient method to ext ract classifying informat ion of hig h dimensional patterns. The key of the method is how to calculate the optimal discriminant vectors. T he stat ist ically uncorrelated opt imal discriminant vectors ensure that the projected features of the pat tern vectors are stat ist ically uncorrelated. The ex isted algorithm comput ing the statistically uncorrelated optimal discriminant vectors is ineffect ive for the case of the small samples ( the w ithin-class scatter matrix is singular) . T his paper presented a unified algorithm that can calculate the stat ist ically uncorrelated opt imal discriminant vectors exactly both for the small samples and large ones. In the case of large samples, the result of our method is the same as the existed method. In order to test the eff iciency of our method, it is used to recognize human faces. Ex perimental result s have show n that our method has bet ter recog nition performance than ex isted method.

参考文献/References:

1  Sammon J W. An optimal discriminant plane. IEEE T rans Computer, 1970, 19( 8) : 826~ 829
2  Foley D H, Sammon J W. An optimal set of discriminant v ectors. IEEE Trans Computer, 1975,24( 3) : 281~ 289
3  Tian Q. Compar ison of stat istical pat tern- recog nit ion algorithms for hybrid processing  : eigenvector-based algorithm. J Opt Soc Am, 1988, A5: 1 670~ 1 672
4  Ho ng Z Q, Yang J Y. Optimal discriminant plane for a small number of samples and desig n method of classifier on t he plane. Pattern Recognition, 1991, 26( 4) : 314~ 317
5Cheng Y Q, Zhuang Y M, Yang J Y. Optimal Fisher discr iminant analysis using the rank decomposition. Patter n Reco gnition, 1992, 25( 1) : 101~ 111
6  Liu K, Cheng Y Q, Yang J Y. Algebr aic feature ex traction fo r imag e recognition based o n an opt-imal discriminant criterio n. Patter n Recog nition, 1993, 26( 8) : 903~ 911
7 Liu K, Cheng Y Q, Yang J Y. An efficient algorithm for Foley- Sammo n optimal set of discriminantvectors by algebraic method. Inter national Journal of Patter n Recognition and Artificial Intell-igence, 1982, 6( 5) : 817~ 829
8Liu K, Cheng Y Q, Yang J Y. A generalized optimal set of discr iminant vectors. Pattern Recogn-ition, 1992, 25( 1) : 731~ 739
9 金忠, 杨静宇, 陆建峰. 一种具有统计不相关性的最佳鉴别矢量集. 计算机学报, 1999,22( 10) : 1 105~ 1 108

备注/Memo

备注/Memo:
国家自然科学基金资助项目 (6 0 0 72 0 34)
束婷婷 女 26 岁 博士生
更新日期/Last Update: 2002-06-30