The Study of On-Line Signature Verification - Using Velocity of Pen-Tip Movement

碩士 === 淡江大學 === 資訊工程學系 === 85 === The signature, as we know, is the unique mark representing of a person. It is usually used in security checking and also is an important evidence which we tell the original from a forgery or whether it i...

Full description

Bibliographic Details
Main Authors: Lee, Pao-Chyee, 李保奇
Other Authors: Lin Hwei-Jen
Format: Others
Language:zh-TW
Published: 1997
Online Access:http://ndltd.ncl.edu.tw/handle/49982446271522018482
Description
Summary:碩士 === 淡江大學 === 資訊工程學系 === 85 === The signature, as we know, is the unique mark representing of a person. It is usually used in security checking and also is an important evidence which we tell the original from a forgery or whether it is a legal paper. Under the circumstances of accomplishing the NII(National Information Infrastructure), document-floating automation is the main issue of the process in which signature verification plays a great role. To achieve this goal, many people have been making their efforts on the research of signature verification. Since a signature can be forged and vary caused by the randomization of the signer, how to extract good features and choose an appropriate classification method becomes the very part of our study. In this paper, an on-line Chinese signature verification system composed of a preprocessing stage and a verification stage is proposed, which uses a dynamic information, the velocity of pen-tip movement while the pen is working on a eletromagnetic tablet, as the only feature for our verification system. Because such a dynamic information does not contain only the characteristics of the signature, but also the particular habit of the signer. First, we break up the trace of the pen-tip movement on the tablet into several primitives through some stages. Further, the average velocity during each segment is separated into two directional components, the X-axis and Y-axis directional components. In the verification stage, we compare the directional components of a testing signature with those of its samples in database by Relaxation Matching and measure Correlation Coefficient as the similarity between them. According to this measurement and the threshold in the database which is measured from the samples by the same signer, we determine whether the signature is real or forgery.