Student: Victor García Romillo
Supervisor: Eva Navas Cordón e Inma Hernáez Rioja
DESCRIPTION
The aim of this project is to generate improvements in speech technologies, especially in the Text-to-Speech field, so that people affected by speech disorders could be benefited by the use of this technology. The current Text-to-Speech developed models are built using DNNs and offer high-quality voices, but still have large margin of improvement. The main objectives to be fulfilled are:
- Improve the robustness of current systems at synthesis time.
- Reduce the dependence on huge amounts of data to train the models.
- Obtain personalized synthetic voices trough improvements over the current systems, especially focusing on personalization from voices with some pathology.