文档介绍：2013 年 7 月四川大学学报(自然科学版) .2013
Jul f
第 50 卷第 4 期( ) .50 .4
Journal of Sichuan University Natural Science Edition Vol No f
:103969/ . .0490‐6756 .2013 .04 .011
doi j issn
基于关键词提取的手写汉字文本
依存笔迹鉴别技术
陈睿,唐雁
(西南大学计算机与信息科学学院,重庆 400715)
摘要:针对文本独立笔迹鉴别技术中需要大篇幅文档以提取图像纹理等笔迹信息的限制条
件,和传统文本依存笔迹鉴别技术中需要内容完全相同的笔迹文档的限制条件,本文提出了一
,对提取出的
相同关键词建立训练集,在两篇文档的相同关键词之间进行匹配,将笔迹鉴别问题转化为签名
,在文档内容不完全相同的情况下,该技术仍然能够有效地对文档进
行笔迹鉴别.
关键词:局部特征;参考表;匹配与投票;笔迹鉴别;签名识别
中图分类号: 391 .1 ; 391 .4 文献标识码: 文章编号:0490‐6756(2013)04‐0719‐09
TP T P A
Writer verification for Chinese handwritten document based on
keywords extraction
C H EN Rui , T A NG Y an
( , , 400715 , )
School puter and Information Science Southwest University Chongqing China
Abstract : ,
In text independent handw riting verification lengthy documents are required to extract image
; ,
textures and other handw riting features w hile in text dependent handw riting verification the contents
. ,
of paring documents are required to be identical In order to e these confinements this
. ,
paper presents a novel text dependent handw riting verification technique In the process keywords are
,
extracted based on matching and voting of local features then same keywords are used to build training
, ,
sets and these train sets of two documents pared so the handw riting verification problem is
.
transformed into signature identification T he experiment results show this method can work effectively
.
under the condition that the contents of two documents are not identical
Key words : , , , , ‐
local feature reference table matching and voting handw ritten verification signature iden
tification
容的文字以鉴别身份,而文本独立需要大篇幅文档
1 ,这两
手写汉字笔迹鉴别可分为文本依存和文本独种条件很难满足[1 ] .针对该问题,本文提出了一种
,在文档内容部分相同
收稿日期:2012‐11‐15
基金项目:中央高校基本科研业务费专项资金( 2012 068) ;教育部春晖计划项目( 2011149)
XDJK C Z
作者简介:陈睿(1979 - ) ,男,硕士,讲师,研究方向为模式