[1]韩宏,杨静宇,娄震.基于层次的分类器组合[J].南京理工大学学报(自然科学版),2002,(01):10-14.
 HanHong YangJingyu LouZhen.Classifier Combination Based on Hierarchical Method[J].Journal of Nanjing University of Science and Technology,2002,(01):10-14.
点击复制

基于层次的分类器组合()
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01期
页码:
10-14
栏目:
出版日期:
2002-02-28

文章信息/Info

Title:
Classifier Combination Based on Hierarchical Method
作者:
韩宏杨静宇娄震
南京理工大学计算机科学与技术系, 南京210094
Author(s):
HanHong YangJingyu LouZhen
Department of Computer Science and Technology,NUST,Nanjing 210094
关键词:
分类器组合 字符识别 特征提取
Keywords:
classif ier combinat ion character recognition feature ext ract ion
分类号:
TP391.4
摘要:
对传统的串行、并行组合方式的分析基础上 ,该文提出一种基于层次的分类器组合方式 ,并以手写体数字识别为例 ,选取了 8个不同信息层次的分类器进行组合 ,得到最终分类结果如下 :正确识别率为 97.15 % ,错误率为 0 .75 % ,拒识率为2 .10 % ,可靠性为 99.2 34%。实验结果表明 ,此种分类器组合方式较传统的串行、并行组合方式在性能上有所提高
Abstract:
Based on the analysis of tradit ional serial and parallel combinat ion method, this paper present s a kind of new classif ier combinat ion based on hierarchical method. This comb-i nat ion method is applied to handwrit ten numeral recog nition. Eight classif iers of different informat ion level are chosen and combined to get the f inal classif icat ion result as follows, correct recognit ion rate 97. 15 % , wrong recognit ion rate 0. 75 %, rejection rate 2. 10 % and reliabil-i ty 99. 234 %. Ex perimental result s show that the classif ier combinat ion method is bet ter than the tradit ional serial and parallel method.

参考文献/References:

1 Woods K, Kegelmeyer W, Bow yer K1 Combination of multiple classifiers using local accur acy est-imates. IEEE Trans PAMI , 1997, 19: 405~ 410
2 Lei Xu1Met hods of combining mult iple classifiers and their applications to handw ritten recogn-ition. IEEE Trans SMC, 1992, 22( 3) : 418~ 435
3 Benediktsson J A, Sw ain P H. Consensus theoretic classification met hods. IEEE Trans SMC, 1992,22( 4) : 688~ 704
4 Liao S X, Pawlak M1On image analysis by moments. IEEE T rans PAMI, 1996, 18( 3) : 254~ 268
5 娄震, 胡钟山, 杨静宇等1 基于轮廓分段特征的手写体阿拉伯数字识别1 计算机学报, 1999,22( 10) : 1 065~ 1 073

备注/Memo

备注/Memo:
国家自然科学基金资助项目(60 0 72 0 34)
韩 宏 女 29 岁 博士生
更新日期/Last Update: 2002-02-28