EXPLORING MODIFICATIONS OF FOURIER TRANSFORM AND THEIR IMPACT ON ACCURACY OF MACHINE LEARNING TECHNIQUES FOR ECG CLASSIFICATION

Aug 5, 2026

Upload by Levi

This master’s thesis explores how modifications of the Fourier transform affect the accuracy of machine learning models for ECG classification. It reviews time-signal classification approaches, preprocessing options, and an ECG dataset framework, including CODE-15. The work analyzes Fourier-transform variants and a proposed FFT-based transformation, builds baseline and modified models, and documents accuracy changes in the frequency domain. Results are then compared against models trained directly on time-domain signals.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages56

Reading Time2 h 21 min

AccessFree to download

Introduction in ECG

Overview of methods for ECG analysis

Overview of dataset CODE-15

Overview of transformations and classifiers in this article

Methodology of experiments

Experimental results

Assignment of master’s thesis Title: Exploring Modiﬁcations of Fourier Transform and their Impact on Accuracy of Machine Learning Techniques for ECG Classiﬁcation Student: Bc. Bogdan Buliakov Supervisor: Ing. Miroslav Čepek, Ph. D. Study program: Informatics Branch / specialization: Knowledge Engineering Department: Department of Applied Mathematics Validity: until the end of summer semester 2024/2025 Instructions This thesis aims to explore machine learning techniques and options to process time signals with focus on ECG. The thesis explores impact of modiﬁcations of Fourier tranform on performace of machine learning models and will compare the accuracy of models in frequency domain with individual modiﬁcations to models working in time domain. Individual steps: 1) Review machine learning models for time signal classiﬁcation (partial and whole) and techniques for preprocessing time signals. 2) Describe and explore suitable dataset. For example CORE-15 for ECG Classiﬁcation. 3) Review Fourier transform, it's modiﬁcations and propose your modiﬁcations. 4) Create a baseline model and experiment with modiﬁed models and document the impact of modiﬁcations on accuracy of the model. 5) Compare the results with techniques directly using the time domain data. Literature: Carlos Mateo, Juan Antonio Talavera. Short-time Fourier transform with the window size ﬁxed in the frequency domain. Digital Signal Processing. Volume 77. 2018. ISSN Electronically approved by Ing. Magda Friedjungová, Ph.D. on 16 February 2024 in Prague. 1051-2004. [https://doi.org/10.1016/j.dsp.2017.1](https://doi.org/10.1016/j.dsp.2017.1)1.003. Ribeiro, Antônio H., et al. "Automatic diagnosis of the 12-lead ECG using a deep neural network." Nature communications 11.1 (2020): 1760. Electronically approved by Ing. Magda Friedjungová, Ph.D. on 16 February 2024 in Prague. Master’s thesis EXPLORING MODIFICATIONS OF FOURIER TRANSFORM AND THEIR IMPACT ON ACCURACY OF MACHINE LEARNING TECHNIQUES FOR ECG CLASSIFICATION Bc. Bogdan Buliakov Faculty of Information Technology Department of Knowledge Engineering Supervisor: Ing. Miroslav ˇCepek Ph.D. May 9, 2024 Czech Technical University in Prague Faculty of Information Technology © 2024 Bc. Bogdan Buliakov. All rights reserved. This thesis is school work as defined by Copyright Act of the Czech Republic. It has been submitted at Czech Technical University in Prague, Faculty of Information Technology. The thesis is protected by the Copyright Act and its usage without author’s permission is prohibited (with exceptions defined by the Copyright Act) . Citation of this thesis: Buliakov Bogdan. Exploring Modifications of Fourier Transform and their Impact on Accuracy of Machine Learning Techniques for ECG Classification. Master’s thesis. Czech Technical University in Prague, Faculty of Information Technology, 2024 . Contents Acknowledgments vi Declaration vii Abstract viii List of abbreviations ix Introduction 1 1 Introduction in ECG 2 2 Overview of methods for ECG analysis 4 2. 1 Classification methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 2.2 Anomaly detection ................................... 6 3 Overview of dataset CODE-15 7 3. 1 Files of dataset . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 4 Overview of transformations and classifiers in this article 10 4.1 My own FFT transformation (FFTwDW) ...................... 10 4.2 Transformations for comparison 13 4.2. 1 Time domain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 4.2.2 Fast Fourier Transform for real values (rFFT) ................ 14 4.2.3 Modified discrete cosine transform (MDCT) ................. 15 4.2.4 Wavelet transform (WT) ........................... 15 4.2.5 Short-Time Fourier Transform with the Window Size Fixed in the Frequency Domain (STFT-FD) ..........................

FAQs

What is the thesis goal regarding Fourier transform modifications?

The thesis aims to evaluate how modifications of the Fourier transform influence the accuracy of machine learning models for ECG classification, comparing frequency-domain models with time-domain approaches.

Which ECG dataset is used in the experiments?

The thesis references the CORE-15 dataset and an overview section for CODE-15 in the contents, indicating its use for ECG classification experiments.

How are modified transforms and classifiers assessed?

A baseline model is built, then models using modified transforms are experimented with; accuracy impacts are documented and compared against models that use time-domain data directly.

Aug 5, 2026

Upload by Levi