Abstract

The work presented in this thesis deals with the development of an on-line Handwritten Signature Verification methodology based on certain Global features that represent the entire signature as a whole. The aim of the research is to come up with an efficient "acceptance/rejection" mechanism that can be used to verify the authenticity of a certain signature. Global features are less expensive to compute than local features, they provide fast verification, and should provide accurate verification results. After identifying the entire global set of attributes that are to be investigated, Cramer's v-statistics and information-theoretic entropy measures of association were used in a dimensionality reduction process to eliminate the attributes that do not contribute significantly to the information of a signature because of their redundant

nature. The resulting sub-set consisted of two binary global qualifiers and nine continuous global attributes. Principal Component Analysis, Hotelling Transform and statistical consistency measures were used to determine the relative weights of the continuous attributes in the verification process. Temporal and dynamic signature features were found to account for about 67. 5% of the total global weights, while purely morphological features represent only about 32.5% of the weight. This result emphasizes the importance of dynamic features in on-line HSV Using the continuous attributes, a weighted Euclidean Distance matching method is used to determine the threshold level for acceptance. By varying this threshold, an inverse relationship was determined between the False Rejection Error (FRE) and False Acceptance Error (FAE). A threshold distance value of0.155 was obtained/or FAE = FRE = 5% at which a Correct Matching Ratio of 95% is achieved using only the 9 continuous global attributes. A complete approach to the problem that also uses the binary qualifiers eliminates FAE and achieves a correct Matching ratio of 97% and a FRE of only 3% at a threshold of 0.165. These results compare favorably with and even are better than current published results, taking into account that the present work presents a very fast acceptance I rejection methodology based on only few global features that produce very close error rates. The processing complexity of this method is very small compared to those using local features and hence, complex matching techniques.

School

School of Sciences and Engineering

Department

Computer Science & Engineering Department

Degree Name

MS in Computer Science

Date of Award

6-1-2003

Online Submission Date

1-1-2003

First Advisor

Amr Goneid

Committee Member 1

Amr Goneid

Committee Member 2

Awad Khalil

Committee Member 3

Amir Zeid

Document Type

Thesis

Extent

102 leaves

Library of Congress Subject Heading 1

Digital signatures

Rights

The American University in Cairo grants authors of theses and dissertations a maximum embargo period of two years from the date of submission, upon request. After the embargo elapses, these documents are made available publicly. If you are the author of this thesis or dissertation, and would like to request an exceptional extension of the embargo period, please write to thesisadmin@aucegypt.edu

Call Number

Thesis 2003/40

Location

mgfth;mrs2

Download

Share

COinS