A machine learning framework for ECG biometric system

• Ano de defesa: 2020
• Autor(a) principal: SANTOS, Alex Barros dos
• Orientador(a): CERQUEIRA, Eduardo Coelho
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal do Pará
• Programa de Pós-Graduação: Programa de Pós-Graduação em Engenharia Elétrica
• Departamento: Instituto de Tecnologia
• País: Brasil
• Palavras-chave em Português: Biometric; Machine Learning; Electrocardiogram; Computer Networks; Wearables
• Área do conhecimento CNPq: CNPQ::ENGENHARIAS::ENGENHARIA ELETRICA
• Link de acesso: https://repositorio.ufpa.br/jspui/handle/2011/17275

The new environment of IoT and the deployment of 5G networks have been generating a huge amount of data. Developers are creating new applications and redesigning other ones completely. Also, a society greater concern with health increases the demand for health services provided with the usage of wearable devices that are getting cheaper. Moreover, the applications require more data protection and privacy. Thus, biometrics has become one of the primary mechanisms for protecting information used by users in all kind of systems and applications. This work investigates the use of an ECG signal in biometrics systems approaching machine learning techniques. This signal is a new alternative not only to increase current safety standards by providing the individual’s continuous authentication but also to assess health with cardiac monitoring already well established in medicine by evaluations. In this context, this master’s thesis proposes some processing steps to data sets, improving its quality that allows it to be used as a reliable source of biometric data. We define techniques for extracting signal considering mobile application constraints and design a structure that allows the use of ECG as a biometric signal in a scalable and heterogeneous environment considering different machine learning techniques such as Support Vector Machine, Random Forest and Neural Networks. The set of our proposed feature extraction, processing steps of data set and a machine learning model are the main contributions of this work.