ECG Biometrics using Deep Neural Networks
Autor(a) principal: | |
---|---|
Data de Publicação: | 2019 |
Tipo de documento: | Dissertação |
Idioma: | eng |
Título da fonte: | Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos) |
Texto Completo: | http://hdl.handle.net/10362/75491 |
Resumo: | Biometrics is a rapidly growing field, with applications in personal identification and security. The Electrocardiogram (ECG) has the potential to be used as a physiological signature for biometric systems. However, current methods still lack in performance across different recording sessions. In this thesis, it is shown that Deep Learning can be applied successfully in the analysis of physiological signals for biometric purposes. This is accomplished in two different experiments by formulating novel approaches based on Convolutional Neural Networks and Recurrent Neural Networks, which may receive heartbeats, signal segments or spectrograms as input. These methods are compared in tasks implying the recognition of subjects from four public databases: Fantasia, ECG-ID, MIT-BIH and CYBHi. This work obtained state-of-the-art results for across-session authentication tasks on the CYBHi dataset, reaching Equal Error Rates of 10.57% and 10.01% for the best model, with corresponding identification accuracy rates of 55.58% and 58.91%. It also demonstrates that using spectrograms as input to the classifier is a promising approach for biometric identification, achieving accuracy values of 99.79% and 96.88%, respectively for Fantasia and ECG-ID databases. Further, it is shown empirically that for ECG biometric systems, the ability of a model to generalize is crucial, not only its capacity to relate and store information. These contributions represent another step towards real-world application of ECGbased biometric systems, closing the gap between intra and inter-session performance and providing some guidelines that can be applied in future work. |
id |
RCAP_17ec74003aa1d3623b5ea618532c8628 |
---|---|
oai_identifier_str |
oai:run.unl.pt:10362/75491 |
network_acronym_str |
RCAP |
network_name_str |
Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos) |
repository_id_str |
7160 |
spelling |
ECG Biometrics using Deep Neural NetworksBiometricsDeep LearningSignal ProcessingElectrocardiogramDomínio/Área Científica::Engenharia e Tecnologia::Engenharia MédicaBiometrics is a rapidly growing field, with applications in personal identification and security. The Electrocardiogram (ECG) has the potential to be used as a physiological signature for biometric systems. However, current methods still lack in performance across different recording sessions. In this thesis, it is shown that Deep Learning can be applied successfully in the analysis of physiological signals for biometric purposes. This is accomplished in two different experiments by formulating novel approaches based on Convolutional Neural Networks and Recurrent Neural Networks, which may receive heartbeats, signal segments or spectrograms as input. These methods are compared in tasks implying the recognition of subjects from four public databases: Fantasia, ECG-ID, MIT-BIH and CYBHi. This work obtained state-of-the-art results for across-session authentication tasks on the CYBHi dataset, reaching Equal Error Rates of 10.57% and 10.01% for the best model, with corresponding identification accuracy rates of 55.58% and 58.91%. It also demonstrates that using spectrograms as input to the classifier is a promising approach for biometric identification, achieving accuracy values of 99.79% and 96.88%, respectively for Fantasia and ECG-ID databases. Further, it is shown empirically that for ECG biometric systems, the ability of a model to generalize is crucial, not only its capacity to relate and store information. These contributions represent another step towards real-world application of ECGbased biometric systems, closing the gap between intra and inter-session performance and providing some guidelines that can be applied in future work.Gamboa, HugoRUNBento, Nuno Filipe Abalada do Val2019-07-15T09:52:16Z2019-0520192019-05-01T00:00:00Zinfo:eu-repo/semantics/publishedVersioninfo:eu-repo/semantics/masterThesisapplication/pdfhttp://hdl.handle.net/10362/75491enginfo:eu-repo/semantics/openAccessreponame:Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)instname:Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informaçãoinstacron:RCAAP2024-03-11T04:34:31Zoai:run.unl.pt:10362/75491Portal AgregadorONGhttps://www.rcaap.pt/oai/openaireopendoar:71602024-03-20T03:35:31.013129Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos) - Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informaçãofalse |
dc.title.none.fl_str_mv |
ECG Biometrics using Deep Neural Networks |
title |
ECG Biometrics using Deep Neural Networks |
spellingShingle |
ECG Biometrics using Deep Neural Networks Bento, Nuno Filipe Abalada do Val Biometrics Deep Learning Signal Processing Electrocardiogram Domínio/Área Científica::Engenharia e Tecnologia::Engenharia Médica |
title_short |
ECG Biometrics using Deep Neural Networks |
title_full |
ECG Biometrics using Deep Neural Networks |
title_fullStr |
ECG Biometrics using Deep Neural Networks |
title_full_unstemmed |
ECG Biometrics using Deep Neural Networks |
title_sort |
ECG Biometrics using Deep Neural Networks |
author |
Bento, Nuno Filipe Abalada do Val |
author_facet |
Bento, Nuno Filipe Abalada do Val |
author_role |
author |
dc.contributor.none.fl_str_mv |
Gamboa, Hugo RUN |
dc.contributor.author.fl_str_mv |
Bento, Nuno Filipe Abalada do Val |
dc.subject.por.fl_str_mv |
Biometrics Deep Learning Signal Processing Electrocardiogram Domínio/Área Científica::Engenharia e Tecnologia::Engenharia Médica |
topic |
Biometrics Deep Learning Signal Processing Electrocardiogram Domínio/Área Científica::Engenharia e Tecnologia::Engenharia Médica |
description |
Biometrics is a rapidly growing field, with applications in personal identification and security. The Electrocardiogram (ECG) has the potential to be used as a physiological signature for biometric systems. However, current methods still lack in performance across different recording sessions. In this thesis, it is shown that Deep Learning can be applied successfully in the analysis of physiological signals for biometric purposes. This is accomplished in two different experiments by formulating novel approaches based on Convolutional Neural Networks and Recurrent Neural Networks, which may receive heartbeats, signal segments or spectrograms as input. These methods are compared in tasks implying the recognition of subjects from four public databases: Fantasia, ECG-ID, MIT-BIH and CYBHi. This work obtained state-of-the-art results for across-session authentication tasks on the CYBHi dataset, reaching Equal Error Rates of 10.57% and 10.01% for the best model, with corresponding identification accuracy rates of 55.58% and 58.91%. It also demonstrates that using spectrograms as input to the classifier is a promising approach for biometric identification, achieving accuracy values of 99.79% and 96.88%, respectively for Fantasia and ECG-ID databases. Further, it is shown empirically that for ECG biometric systems, the ability of a model to generalize is crucial, not only its capacity to relate and store information. These contributions represent another step towards real-world application of ECGbased biometric systems, closing the gap between intra and inter-session performance and providing some guidelines that can be applied in future work. |
publishDate |
2019 |
dc.date.none.fl_str_mv |
2019-07-15T09:52:16Z 2019-05 2019 2019-05-01T00:00:00Z |
dc.type.status.fl_str_mv |
info:eu-repo/semantics/publishedVersion |
dc.type.driver.fl_str_mv |
info:eu-repo/semantics/masterThesis |
format |
masterThesis |
status_str |
publishedVersion |
dc.identifier.uri.fl_str_mv |
http://hdl.handle.net/10362/75491 |
url |
http://hdl.handle.net/10362/75491 |
dc.language.iso.fl_str_mv |
eng |
language |
eng |
dc.rights.driver.fl_str_mv |
info:eu-repo/semantics/openAccess |
eu_rights_str_mv |
openAccess |
dc.format.none.fl_str_mv |
application/pdf |
dc.source.none.fl_str_mv |
reponame:Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos) instname:Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informação instacron:RCAAP |
instname_str |
Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informação |
instacron_str |
RCAAP |
institution |
RCAAP |
reponame_str |
Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos) |
collection |
Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos) |
repository.name.fl_str_mv |
Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos) - Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informação |
repository.mail.fl_str_mv |
|
_version_ |
1799137976132829184 |