Home Research Speech Technology
Conversor Texto-Voz PDF Print

Welcome to www Text-to-speech Activity Report of G.A.P.S.'s page. We are a research group into S.S.R. (Signals, Systems and Radiocommunications) inside of U.P.M.(Universidad Politécnica de Madrid) (Spain)


MAJOR RESEARCH ACTIVITY ON TEXT-TO-SPEECH

  • Improve naturalness in state-of-the-art TTS systems

  • Joint F0 and duration prosodic modeling

  • TTS adaptation to specific speakers and tasks

  • Fully automatic prosodic modeling from task or speaker dependent text corpora and speech material.

We are members of the COST 258 Action on "The Naturalness of Synthetic Speech".

 

The main publications are accesible at HERE

A preliminary evaluation has been made over a population of 12 people . A set of 20 sentences were synthesized for both applications using three different prosodic models. For further details please review [Lopez-Gonzalo et al., 1997] PS bibtex

 

AUDIO DEMONSTRATION:

 

1. Model A: general purpose prosodic model.

Example 1: Download Embed Embed this video on your site

Example 2: Download Embed Embed this video on your site

Example 3: Download Embed Embed this video on your site

2. Model B: modelo prosódico adaptado al locutor y a la tarea con categorias gramaticales generales.

Example 1: Download Embed Embed this video on your site

Example 2: Download Embed Embed this video on your site

Example 3: Download Embed Embed this video on your site

3. Model C: modelo prosódico adaptado al locutor y a la tarea con categorias gramaticales generales

Example 1: Download Embed Embed this video on your site

Example 2: Download Embed Embed this video on your site

Example 3: Download Embed Embed this video on your site

 

The TTS system used for the synthetic speech is an experimental diphone-based Synthesizer developed at ETSI Telecomunicación - Universidad de Vigo (Spain). An evolution of this system is at  COTOVIA TTS ETSI Telecomunicación - Universidad de Vigo


Last Updated on Thursday, 01 October 2009 18:39