Writer identification based on graphology techniques

Author

Santana, Omar ; Travieso, Carlos M. ; Alonso, Jesús B. ; Ferrer, Miguel A.

Author_Institution

Dipt. de Senales y Comun., Univ. de Las Palmas de Gran Canaria, Las Palmas de Gran Canaria

fYear

2008

fDate

13-16 Oct. 2008

Firstpage

167

Lastpage

173

Abstract

In this work an innovative system biometric of writerspsila identification based in technical expert calligraphic and graphology on the handwritten script is presented. It has been developed working in off-line mode on a Spanish words image database, formed by 29 different individuals. All the extractions of characteristics carried out on the images and that they have been used for the identification, they were carried out by means of the estimate of several elements object using studies of The French Graphology School, and that they are commonly employed by the handwriting experts in judicial matters. The success percentage achieved with 5 of these characteristics from this database of 29 writers is of 99.34%. In new experimentations, with these same parameters and enlarging the database to 70 users, a rate of success of 92% was reached.

Keywords

handwritten character recognition; neural nets; support vector machines; Spanish word image database; artificial neural network; biometric identification; graphology technique; handwritten script; handwritten writer recognition; support vector machine; technical expert calligraphic; Artificial neural networks; Biometrics; Handwriting recognition; Image databases; Law; Pins; Robustness; Security; Support vector machines; Writing; Artificial Neural Networks (NNs) and Support Vector Machines (SVMs); Biometric Identification; Graphology Techniques; Writer Identification;

fLanguage

English

Publisher

ieee

Conference_Titel

Security Technology, 2008. ICCST 2008. 42nd Annual IEEE International Carnahan Conference on

Conference_Location

Prague

Print_ISBN

978-1-4244-1816-9

Electronic_ISBN

978-1-4244-1817-6

Type

conf

DOI

10.1109/CCST.2008.4751297

Filename

4751297