Your browser does not support JavaScript!

Home    Search  

Results - Details

Search command : Author="Παπαδάκη"  And Author="Αικατερίνη"

Current Record: 4 of 13

Back to Results Previous page
Next page
Add to Basket
[Add to Basket]
Identifier 000433804
Title Modeling of speech signals using recurrent neural networks
Alternative Title Μοντελοποίηση σημάτων φωνής με την χρήση αναδρομικών νευρωνικών
Author Παπαδάκη, Αικατερίνη Ι
Thesis advisor Κομίνης, Ιωάννης
Πανταζής, Γιάννης
Reviewer Μακρής, Κωνσταντίνος
Abstract Speech is the primary means of communication between humans. A speech signal waveform corresponds to the variation of air pressure over time. Over two hundred years ago there have been early efforts to produce synthetic speech, using mechanical apparatus. Recently, ongoing activities have been focused on the artificial production of human speech, or else speech synthesis. Particularly, there have been developed vocoders, analyzers and synthesizers of human voice signals, based on mathematical models. Furthermore, neural vocoders, vocoders based on artificial neural networks, have introduced a new area in speech synthesis. Goal of this study is to build a neural vocoder which combines deep learning with prior knowledge of the inherent sinusoidal nature of speech, in contrast with the majority of available neural based vocoders which discard signal processing in favor of neural networks. Particularly, the origin of the present study is modeling complex multi-component AM and FM sinusoidal waves with the property to represent speech signals, employing Recurrent Neural Networks (RNNs). The target is to develop a light neural vocoder, faster than WaveRNN, which achieves state-of-the-art performance. In the context of this thesis, we implement a variant of the WaveRNN model and we present the generated state-of-the-art results. Furthermore, we examine the proposed model’s performance using synthetic as well as real speech signals.
Language English
Subject Neural vocoder
Recurrent neural networks
Sinusoidal signals
Αναδρομικά νευρωνικά δύκτια
Ημιτονοειδή σήματα
Νευρονικός Φωνοκωδικοποιητής
Νευρωνικά δύκτια
Σήματα φωνής
Issue date 2021-03-24
Collection   School/Department--School of Sciences and Engineering--Department of Physics--Post-graduate theses
  Type of Work--Post-graduate theses
Permanent Link https://elocus.lib.uoc.gr//dlib/7/d/5/metadata-dlib-1604663338-495972-29058.tkl Bookmark and Share
Views 395

Digital Documents
No preview available

Download document
View document
Views : 9