Multilingual Speech Synthesis with Estonian
Name
Rasmus Lellep
Abstract
The need for speech synthesis, including multilingual synthesis is continuously growing. While earlier approaches of speech synthesis were based on manually composed linguistic features or imitating the components of human speech system, in the recent years there have been a lot of attempts of creating solutions that require less preparation and are less demanding. These solutions are largely based on machine learning, primarily artificial neural networks. This thesis gives an overview to some of the afore-mentioned solutions, which also focus on multilingual synthesis. None of the multilingual solutions have the support for Estonian. In the practical part of this thesis, the first multilingual text-to-speech synthesizer with support for Estonian was made. The synthesizer works with 11 languages in total. The thesis describes the steps of creating this synthesizer. With composed test sentences, the synthesizer's Estonian speech quality was evaluated. Finally, some adjustments that could further improve the Estonian speech quality were proposed.
Graduation Thesis language
Estonian
Graduation Thesis type
Bachelor - Computer Science
Supervisor(s)
Liisa Rätsep
Defence year
2021