Original title: EKG biometrie s využitím hlubokého učení
Translated title: ECG biometrics using deep learning
Authors: Repčík, Tomáš ; Mézl, Martin (referee) ; Vičar, Tomáš (advisor)
Document type: Master’s theses
Year: 2021
Language: slo
Publisher: Vysoké učení technické v Brně. Fakulta elektrotechniky a komunikačních technologií
Abstract: [slo] [eng]
Diplomová práca spracováva celkový prehľad rôznych prístupov k využitiu EKG ako biometrie. Pre trénovanie a testovanie vytvorených neurónových sietí boli zozbierané záznamy ľudí pomocou náramku schopného merať EKG v priebehu štyroch mesiacov. Pomocou týchto dát boli zostrojené neurónové siete rôznych typov, z ktorých najlepšie obstáli jednoduché konvolučné neurónové siete. Modely v priemere dosahovali skutočnej miery prijatia nad 98,9% a odmietnutia 99,5%. Následne po tréningu boli vizualizované rôznymi technikami a popísane dôležité úseky EKG pre verifikáciu. Práca taktiež popisuje implementáciu v podobe Android aplikácie a optimalizáciu neurónových sietí. This diploma thesis makes a comprehensive overview of various approaches to the usage of ECG as biometry. ECG of people was measured with ECG capable wristband for training and testing purposes. The measurements were gathered during the four-month period. The neural networks of various types were trained on these data, and the feedforward convolutional neural networks have the best performance. These models reached a true acceptance rate 98,9% and a true rejection rate 99,5% on average. After training, the models have been visualised with a variety of techniques and essentials parts of ECG for verification have been described. The thesis also describes the first implementation of the Android application.
Keywords: Android application; biometry; convolutional neural networks; Deep learning; ECG; neural networks; siamese neural networks; visualisation of neural networks

Institution: Brno University of Technology (web)
Document availability information: Fulltext is available in the Brno University of Technology Digital Library.
Original record: https://hdl.handle.net/11012/246157

Permalink: http://www.nusl.cz/ntk/nusl-617265


The record appears in these collections:
Universities and colleges > Public universities > Brno University of Technology
Academic theses (ETDs) > Master’s theses