权益分类	功能权益	普通用户	{{item.name}}会员
{{category.name}}	{{benefitItem.name}}

Automatic Verification of Handwritten Signature

自动验证手写签名

基本信息

批准号：
08650513
负责人：
MATOZAKI Takeshi
金额：
$ 1.22万
依托单位：
Science University of Tokyo
依托单位国家：
日本
项目类别：
Grant-in-Aid for Scientific Research (C)
财政年份：
1996
资助国家：
日本
起止时间：
1996 至 1997
项目状态：
已结题

项目摘要

Automatic recognition of the individual is a big research theme on advances of the information society. We pointed out that there are personal characteristics in a writer's calligraphy of the signature.In the recognition of individual, important aims are as follows.1) In a calligraphy of handwriting, founding out the charactes which the individual characteristics are included.2 Discrimination of a writer's calligraphy of handwriting using the characteristics, effectively.In this research, we extracted some kind of personal characteristics from an alphabet signature in offline. 1st of all, we assured usefulness of the characteristics using a vagueness degree function, individually. Next, by combination of these useful characteristics, we reduce the vagueness degree of discrimination. By the way, we introduced the order information of writing as a new charactristics.We like to summarize as follws.1) In combination with characteristics which minimize the vagueness degree of discrimination for an individual writer's calligraphy, we realized discriminant factor 91.7%2) In order to level up total efficiency through the discrimination system, we selected the characteristics of discrimination in common each signature.In this case, the discriminant factor is 88.5% and the discriminant time isabout 1 sec. And also, to this offline discrimination of signature, introduction of the write order information is new trial, usefulness of the trial was clarified. This is worthy of special mention.

个人的自动识别是信息社会进步的一个大的研究主题。作者指出，签名笔迹中存在着个人特征，在对签名笔迹的识别中，重要的目的是：1）在签名笔迹中，找出包含个人特征的特征; 2）利用签名笔迹中的特征，有效地鉴别签名笔迹。我们离线从一个字母签名中提取了某种个人特征。首先，我们单独地使用相似度函数来确保特征的有用性。其次，通过结合这些有用的特征，我们降低了歧视的严重程度。本文的主要工作如下：1）结合对单个书法家的识别准确度要求最低的特征，实现了91.7%的识别率; 2）为了提高识别系统的整体效率，我们选取了每个特征的共同鉴别特征，在这种情况下，鉴别因子为88.5%，鉴别时间约为1秒。此外，对于签名的离线识别，写入命令信息的引入是新的尝试，阐明了该尝试的有用性。这一点特别值得一提。