Writer-independent offline handwritten signature verification system based on the dichotomy transformation, prototype selection and feature selection
Autor(a) principal: | |
---|---|
Data de Publicação: | 2020 |
Tipo de documento: | Tese |
Idioma: | eng |
Título da fonte: | Repositório Institucional da UFPE |
Texto Completo: | https://repositorio.ufpe.br/handle/123456789/38547 |
Resumo: | High number of writers, small number of training samples per writer with high intraclass variability and heavily imbalanced class distributions are among the challenges and difficulties of the offline Handwritten Signature Verification (HSV) problem. A good alternative to tackle these issues is to use a writer-independent (WI) framework. In WI systems, a single model is trained to perform signature verification for all writers from a dissimilarity space generated by the dichotomy transformation. Among the advantages of this framework is its scalability to deal with some of these challenges and its ease in managing new writers, and hence of being used in a transfer learning context. In this work, a deep analysis of this approach is presented, highlighting how it handles the challenges as well as the dynamic selection of reference signatures through fusion function, and its application for transfer learning. All the analyses are carried out using the instance hardness (IH) measure. By having these findings at the instance level, we develop an approach that uses prototype selection (Condensed Nearest Neighbors) and feature selection (based on Binary Particle Swarm Optimization) techniques well suited to our WI-HSV scenario. These techniques allowed us to handle the redundancy of information in both sample and the feature levels present in the dissimilarity space. Specifically in the feature selection scenario, we also propose a global validation strategy with an external archive to control overfitting during the search process. The experimental results reported herein show that the use of prototype selection and feature selection in the dissimilarity space allows a reduction in its redundant information and the complexity of the classifier without degrading its generalization performance. In addition, the results show that the WI classifier is scalable enough to be used in a transfer learning approach, with a resulting performance comparable to that of a classifier trained and tested in the same dataset. Finally, using the IH analysis, we were able to characterize “good” and “bad” quality skilled forgeries as well as the frontier region between positive and negative samples. |
id |
UFPE_e7546019c0412d547c60b288939992a5 |
---|---|
oai_identifier_str |
oai:repositorio.ufpe.br:123456789/38547 |
network_acronym_str |
UFPE |
network_name_str |
Repositório Institucional da UFPE |
repository_id_str |
2221 |
spelling |
SOUZA, Victor Lorena de Fariashttp://lattes.cnpq.br/0018643790738396http://lattes.cnpq.br/5194381227316437http://lattes.cnpq.br/6269525393139517http://lattes.cnpq.br/1143656271684404OLIVEIRA, Adriano Lorena Inacio deSABOURIN, RobertCRUZ, Rafael Menelau Oliveira e2020-11-09T18:10:45Z2020-11-09T18:10:45Z2020-08-12SOUZA, Victor Lorena de Farias. Writer-independent offline handwritten signature verification system based on the dichotomy transformation, prototype selection and feature selection. 2020. Tese (Doutorado em Ciência da Computação) – Universidade Federal de Pernambuco, Recife, 2020.https://repositorio.ufpe.br/handle/123456789/38547High number of writers, small number of training samples per writer with high intraclass variability and heavily imbalanced class distributions are among the challenges and difficulties of the offline Handwritten Signature Verification (HSV) problem. A good alternative to tackle these issues is to use a writer-independent (WI) framework. In WI systems, a single model is trained to perform signature verification for all writers from a dissimilarity space generated by the dichotomy transformation. Among the advantages of this framework is its scalability to deal with some of these challenges and its ease in managing new writers, and hence of being used in a transfer learning context. In this work, a deep analysis of this approach is presented, highlighting how it handles the challenges as well as the dynamic selection of reference signatures through fusion function, and its application for transfer learning. All the analyses are carried out using the instance hardness (IH) measure. By having these findings at the instance level, we develop an approach that uses prototype selection (Condensed Nearest Neighbors) and feature selection (based on Binary Particle Swarm Optimization) techniques well suited to our WI-HSV scenario. These techniques allowed us to handle the redundancy of information in both sample and the feature levels present in the dissimilarity space. Specifically in the feature selection scenario, we also propose a global validation strategy with an external archive to control overfitting during the search process. The experimental results reported herein show that the use of prototype selection and feature selection in the dissimilarity space allows a reduction in its redundant information and the complexity of the classifier without degrading its generalization performance. In addition, the results show that the WI classifier is scalable enough to be used in a transfer learning approach, with a resulting performance comparable to that of a classifier trained and tested in the same dataset. Finally, using the IH analysis, we were able to characterize “good” and “bad” quality skilled forgeries as well as the frontier region between positive and negative samples.FACEPEGrande número de escritores, poucas amostras de treinamento por escritor, com alta variabilidade intra-classe e distribuições de classes fortemente desequilibradas, estão entre os desafios e as dificuldades da Verificação de Assinatura Manuscrita (HSV) offline. Uma boa alternativa para resolver esses problemas é usar um método independente de escritor (WI). Nos sistemas WI, um único modelo de classificação é treinado para executar a verificação de assinatura de todos os escritores a partir de um espaço de dissimilaridade gerado pela transformação dicotômica. Entre as vantagens dessa estrutura estão: a escalabilidade para lidar com alguns desses desafios listados e a facilidade no gerenciamento de novos escritores, e, portanto, a sua utilização em um contexto de transferência de aprendizado. Neste trabalho, apresentamos uma análise aprofundada dessa abordagem, destacando como ela lida com os desafios, a seleção dinâmica de assinaturas de referência por meio da função de fusão e sua aplicação na transferência de aprendizado. Todas as análises são realizadas usando a medida de dificuldade da instância (IH). Tendo por base os resultados dessas análises, desenvolvemos uma abordagem que usa técnicas de seleção de protótipos (vizinhos mais próximos condensados) e de seleção de características (com base na otimização de enxame de partículas binárias) adequadas ao nosso cenário WI-HSV. Essas técnicas nos permitiram lidar com a redundância de informações nos níveis das amostras e das características presentes no espaço de dissimilaridades. Especificamente no cenário de seleção de características, também propomos uma estratégia de validação global com um arquivo externo para controlar o overfitting durante o processo de busca. Os resultados experimentais relatados aqui mostram que o uso da seleção de protótipos e seleção de características no espaço de dissimilaridade permite uma redução em suas informações redundantes e na complexidade do classificador sem degradar seu desempenho de generalização. Além disso, os resultados mostram que o classificador WI é escalável o suficiente para ser usado em uma abordagem de aprendizado de transferência, com um desempenho resultante comparável ao de um classificador treinado e testado no mesmo conjunto de dados. Por fim, os resultados experimentais mostram que, utilizando a análise IH, conseguimos caracterizar falsificações especializadas de qualidade “boa” e “ruim”, bem como a região fronteiriça entre amostras positivas e negativas.engUniversidade Federal de PernambucoPrograma de Pos Graduacao em Ciencia da ComputacaoUFPEBrasilAttribution-NonCommercial-NoDerivs 3.0 Brazilhttp://creativecommons.org/licenses/by-nc-nd/3.0/br/info:eu-repo/semantics/openAccessInteligência computacionalTransformação dicotômicaWriter-independent offline handwritten signature verification system based on the dichotomy transformation, prototype selection and feature selectioninfo:eu-repo/semantics/publishedVersioninfo:eu-repo/semantics/doctoralThesisdoutoradoreponame:Repositório Institucional da UFPEinstname:Universidade Federal de Pernambuco (UFPE)instacron:UFPECC-LICENSElicense_rdflicense_rdfapplication/rdf+xml; charset=utf-8811https://repositorio.ufpe.br/bitstream/123456789/38547/2/license_rdfe39d27027a6cc9cb039ad269a5db8e34MD52ORIGINALTESE Victor Lorena de Farias Souza.pdfTESE Victor Lorena de Farias Souza.pdfapplication/pdf5577948https://repositorio.ufpe.br/bitstream/123456789/38547/1/TESE%20Victor%20Lorena%20de%20Farias%20Souza.pdfcd618433d6fca0689e114fade219a08aMD51TEXTTESE Victor Lorena de Farias Souza.pdf.txtTESE Victor Lorena de Farias Souza.pdf.txtExtracted texttext/plain240505https://repositorio.ufpe.br/bitstream/123456789/38547/4/TESE%20Victor%20Lorena%20de%20Farias%20Souza.pdf.txt5e574e01c3c5f2ef24ce4faef5d85e6dMD54THUMBNAILTESE Victor Lorena de Farias Souza.pdf.jpgTESE Victor Lorena de Farias Souza.pdf.jpgGenerated Thumbnailimage/jpeg1286https://repositorio.ufpe.br/bitstream/123456789/38547/5/TESE%20Victor%20Lorena%20de%20Farias%20Souza.pdf.jpgd94483d1125a08d5d2b47b342492a6c3MD55LICENSElicense.txtlicense.txttext/plain; charset=utf-82310https://repositorio.ufpe.br/bitstream/123456789/38547/3/license.txtbd573a5ca8288eb7272482765f819534MD53123456789/385472020-11-10 02:16:32.791oai:repositorio.ufpe.br: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ório InstitucionalPUBhttps://repositorio.ufpe.br/oai/requestattena@ufpe.bropendoar:22212020-11-10T05:16:32Repositório Institucional da UFPE - Universidade Federal de Pernambuco (UFPE)false |
dc.title.pt_BR.fl_str_mv |
Writer-independent offline handwritten signature verification system based on the dichotomy transformation, prototype selection and feature selection |
title |
Writer-independent offline handwritten signature verification system based on the dichotomy transformation, prototype selection and feature selection |
spellingShingle |
Writer-independent offline handwritten signature verification system based on the dichotomy transformation, prototype selection and feature selection SOUZA, Victor Lorena de Farias Inteligência computacional Transformação dicotômica |
title_short |
Writer-independent offline handwritten signature verification system based on the dichotomy transformation, prototype selection and feature selection |
title_full |
Writer-independent offline handwritten signature verification system based on the dichotomy transformation, prototype selection and feature selection |
title_fullStr |
Writer-independent offline handwritten signature verification system based on the dichotomy transformation, prototype selection and feature selection |
title_full_unstemmed |
Writer-independent offline handwritten signature verification system based on the dichotomy transformation, prototype selection and feature selection |
title_sort |
Writer-independent offline handwritten signature verification system based on the dichotomy transformation, prototype selection and feature selection |
author |
SOUZA, Victor Lorena de Farias |
author_facet |
SOUZA, Victor Lorena de Farias |
author_role |
author |
dc.contributor.authorLattes.pt_BR.fl_str_mv |
http://lattes.cnpq.br/0018643790738396 |
dc.contributor.advisorLattes.pt_BR.fl_str_mv |
http://lattes.cnpq.br/5194381227316437 |
dc.contributor.advisor-coLattes.pt_BR.fl_str_mv |
http://lattes.cnpq.br/6269525393139517 http://lattes.cnpq.br/1143656271684404 |
dc.contributor.author.fl_str_mv |
SOUZA, Victor Lorena de Farias |
dc.contributor.advisor1.fl_str_mv |
OLIVEIRA, Adriano Lorena Inacio de |
dc.contributor.advisor-co1.fl_str_mv |
SABOURIN, Robert CRUZ, Rafael Menelau Oliveira e |
contributor_str_mv |
OLIVEIRA, Adriano Lorena Inacio de SABOURIN, Robert CRUZ, Rafael Menelau Oliveira e |
dc.subject.por.fl_str_mv |
Inteligência computacional Transformação dicotômica |
topic |
Inteligência computacional Transformação dicotômica |
description |
High number of writers, small number of training samples per writer with high intraclass variability and heavily imbalanced class distributions are among the challenges and difficulties of the offline Handwritten Signature Verification (HSV) problem. A good alternative to tackle these issues is to use a writer-independent (WI) framework. In WI systems, a single model is trained to perform signature verification for all writers from a dissimilarity space generated by the dichotomy transformation. Among the advantages of this framework is its scalability to deal with some of these challenges and its ease in managing new writers, and hence of being used in a transfer learning context. In this work, a deep analysis of this approach is presented, highlighting how it handles the challenges as well as the dynamic selection of reference signatures through fusion function, and its application for transfer learning. All the analyses are carried out using the instance hardness (IH) measure. By having these findings at the instance level, we develop an approach that uses prototype selection (Condensed Nearest Neighbors) and feature selection (based on Binary Particle Swarm Optimization) techniques well suited to our WI-HSV scenario. These techniques allowed us to handle the redundancy of information in both sample and the feature levels present in the dissimilarity space. Specifically in the feature selection scenario, we also propose a global validation strategy with an external archive to control overfitting during the search process. The experimental results reported herein show that the use of prototype selection and feature selection in the dissimilarity space allows a reduction in its redundant information and the complexity of the classifier without degrading its generalization performance. In addition, the results show that the WI classifier is scalable enough to be used in a transfer learning approach, with a resulting performance comparable to that of a classifier trained and tested in the same dataset. Finally, using the IH analysis, we were able to characterize “good” and “bad” quality skilled forgeries as well as the frontier region between positive and negative samples. |
publishDate |
2020 |
dc.date.accessioned.fl_str_mv |
2020-11-09T18:10:45Z |
dc.date.available.fl_str_mv |
2020-11-09T18:10:45Z |
dc.date.issued.fl_str_mv |
2020-08-12 |
dc.type.status.fl_str_mv |
info:eu-repo/semantics/publishedVersion |
dc.type.driver.fl_str_mv |
info:eu-repo/semantics/doctoralThesis |
format |
doctoralThesis |
status_str |
publishedVersion |
dc.identifier.citation.fl_str_mv |
SOUZA, Victor Lorena de Farias. Writer-independent offline handwritten signature verification system based on the dichotomy transformation, prototype selection and feature selection. 2020. Tese (Doutorado em Ciência da Computação) – Universidade Federal de Pernambuco, Recife, 2020. |
dc.identifier.uri.fl_str_mv |
https://repositorio.ufpe.br/handle/123456789/38547 |
identifier_str_mv |
SOUZA, Victor Lorena de Farias. Writer-independent offline handwritten signature verification system based on the dichotomy transformation, prototype selection and feature selection. 2020. Tese (Doutorado em Ciência da Computação) – Universidade Federal de Pernambuco, Recife, 2020. |
url |
https://repositorio.ufpe.br/handle/123456789/38547 |
dc.language.iso.fl_str_mv |
eng |
language |
eng |
dc.rights.driver.fl_str_mv |
Attribution-NonCommercial-NoDerivs 3.0 Brazil http://creativecommons.org/licenses/by-nc-nd/3.0/br/ info:eu-repo/semantics/openAccess |
rights_invalid_str_mv |
Attribution-NonCommercial-NoDerivs 3.0 Brazil http://creativecommons.org/licenses/by-nc-nd/3.0/br/ |
eu_rights_str_mv |
openAccess |
dc.publisher.none.fl_str_mv |
Universidade Federal de Pernambuco |
dc.publisher.program.fl_str_mv |
Programa de Pos Graduacao em Ciencia da Computacao |
dc.publisher.initials.fl_str_mv |
UFPE |
dc.publisher.country.fl_str_mv |
Brasil |
publisher.none.fl_str_mv |
Universidade Federal de Pernambuco |
dc.source.none.fl_str_mv |
reponame:Repositório Institucional da UFPE instname:Universidade Federal de Pernambuco (UFPE) instacron:UFPE |
instname_str |
Universidade Federal de Pernambuco (UFPE) |
instacron_str |
UFPE |
institution |
UFPE |
reponame_str |
Repositório Institucional da UFPE |
collection |
Repositório Institucional da UFPE |
bitstream.url.fl_str_mv |
https://repositorio.ufpe.br/bitstream/123456789/38547/2/license_rdf https://repositorio.ufpe.br/bitstream/123456789/38547/1/TESE%20Victor%20Lorena%20de%20Farias%20Souza.pdf https://repositorio.ufpe.br/bitstream/123456789/38547/4/TESE%20Victor%20Lorena%20de%20Farias%20Souza.pdf.txt https://repositorio.ufpe.br/bitstream/123456789/38547/5/TESE%20Victor%20Lorena%20de%20Farias%20Souza.pdf.jpg https://repositorio.ufpe.br/bitstream/123456789/38547/3/license.txt |
bitstream.checksum.fl_str_mv |
e39d27027a6cc9cb039ad269a5db8e34 cd618433d6fca0689e114fade219a08a 5e574e01c3c5f2ef24ce4faef5d85e6d d94483d1125a08d5d2b47b342492a6c3 bd573a5ca8288eb7272482765f819534 |
bitstream.checksumAlgorithm.fl_str_mv |
MD5 MD5 MD5 MD5 MD5 |
repository.name.fl_str_mv |
Repositório Institucional da UFPE - Universidade Federal de Pernambuco (UFPE) |
repository.mail.fl_str_mv |
attena@ufpe.br |
_version_ |
1802310790362955776 |