|
Identifier |
000460745 |
Title |
Application of convolutional neural networks in inverse problems of Ocean acoustics |
Alternative Title |
Εφαρμογή συνελικτικών νευρωνικών δικτύων σε αντίστροφα προβλήματα θαλάσσιας ακουστικής |
Author
|
Τζιράκης, Βασίλειος
|
Thesis advisor
|
Ταρουδάκης, Μιχαήλ
|
Reviewer
|
Χατζηπαντελής, Παναγιώτης
Μακράκης, Γεώργιος
|
Abstract |
The main goal of the thesis is to study the application of a Convolutional Neural Network for the estimation
of the dispersion curves that characterize the spectrogram of an acoustic signal recorded in an ocean
waveguide. Dispersion curves are input data for a class of inverse problems in underwater acoustics
which are treated using the time frequency analysis of a recorded signal. To make a concise presentation
of my work, the thesis starts with a brief presentation of wave theory with a special reference to modeling
a broadband source. Chapter 2 is devoted to Fourier transform as it is the main tool for signal processing
in my work. The spectrogram is introduced in this chapter. As the applications considered in the thesis
are related to inverse problems, an introduction to the theory of inverse problems is is presented in Chapter
3. In Chapter 4, the basic features of a Convolutional Neural Network (CNN), which is the main tool
used to extract the information on the dispersion curves, are presented. Finally Chapter 5 presents a test
case on the application of the CNN in a simulated underwater acoustic signal, which is of the form of
signals used in applications of ocean acoustic tomography or seabed classification. The conclusions of
my study are presented in the final chapter.
|
Language |
English |
Subject |
Wave theory |
|
Κυματική θεωρία |
Issue date |
2023-11-24 |
Collection
|
School/Department--School of Sciences and Engineering--Department of Mathematics and Applied Mathematics--Post-graduate theses
|
|
Type of Work--Post-graduate theses
|
Permanent Link |
https://elocus.lib.uoc.gr//dlib/d/1/b/metadata-dlib-1701327555-851481-19370.tkl
|
Views |
885 |