John Ullyses O. Velaroso, Sherwin Carlo R. Cruz. 4 0
Enhancement of offline signature verification using support vector machine / 6 6 John Ullyses O. Velaroso, Sherwin Carlo R. Cruz. - - - 122 pp. 28 cm. - - - - - . - . - 0 . - . - 0 .
Undergraduate Thesis: (Bachelor of Science in Computer Science) Pamantasan ng Lungsod ng Maynila, 2022.
5
ABSTRACT: Signatures are commonly used for personal verification or signifying an agreement. Therefore, confirming whether a signature is genuine has become a priority in information security. Signature verification methods are continually developed to automate and hasten the process. A study utilized Support Vector Machine to verify offline signatures. This paper proposed an enhancement of the said algorithm by adding preprocessing techniques, using a more robust set of feature extraction methods, utilizing a writer-dependent classifier, and computing the best hyperparameter values of C and Gamma using the Giza Pyramid Construction Algorithm. Two datasets of offline signatures were used to train and test the algorithms. Results showed that the enhanced algorithm significantly performed better than the existing algorithm, with an accuracy score of 96% and FAR and FRR both having 4% using the CEDAR dataset. The GPDS-300 resulted to an accuracy score of 92%, FAR of 11% and FRR of 6%. In comparison to the initial algorithm that resulted with an accuracy score of 83% and, FRR both having 17% using the CEDAR dataset, and 83% accuracy score as well in the GPDS-300, FAR of 17%, and FRR of 16%.
5
2 = =
2
2 --0------
6 --0-- 2 --------
0 2 --
--20------
--------20--
--------20--
----2
/ 2
/ 2
/
/
Enhancement of offline signature verification using support vector machine / 6 6 John Ullyses O. Velaroso, Sherwin Carlo R. Cruz. - - - 122 pp. 28 cm. - - - - - . - . - 0 . - . - 0 .
Undergraduate Thesis: (Bachelor of Science in Computer Science) Pamantasan ng Lungsod ng Maynila, 2022.
5
ABSTRACT: Signatures are commonly used for personal verification or signifying an agreement. Therefore, confirming whether a signature is genuine has become a priority in information security. Signature verification methods are continually developed to automate and hasten the process. A study utilized Support Vector Machine to verify offline signatures. This paper proposed an enhancement of the said algorithm by adding preprocessing techniques, using a more robust set of feature extraction methods, utilizing a writer-dependent classifier, and computing the best hyperparameter values of C and Gamma using the Giza Pyramid Construction Algorithm. Two datasets of offline signatures were used to train and test the algorithms. Results showed that the enhanced algorithm significantly performed better than the existing algorithm, with an accuracy score of 96% and FAR and FRR both having 4% using the CEDAR dataset. The GPDS-300 resulted to an accuracy score of 92%, FAR of 11% and FRR of 6%. In comparison to the initial algorithm that resulted with an accuracy score of 83% and, FRR both having 17% using the CEDAR dataset, and 83% accuracy score as well in the GPDS-300, FAR of 17%, and FRR of 16%.
5
2 = =
2
2 --0------
6 --0-- 2 --------
0 2 --
--20------
--------20--
--------20--
----2
/ 2
/ 2
/
/