Personal tools
You are here: Home
Navigation
Log in


Forgot your password?
New user?
August 2014 »
Su Mo Tu We Th Fr Sa
1 2
3 4 5 6 7 8 9
10 11 12 13 14 15 16
17 18 19 20 21 22 23
24 25 26 27 28 29 30
31
 
Document Actions

Speech Synthesis Group

by spy last modified 2010-11-05 18:46

Welcome to the website of the Speech Synthesis Group.

The Speech Synthesis Group at the Institute for Language and Speech Processing (ILSP) of the "Athena" Research Center is one of the oldest groups in speech synthesis in Greece. It has been in the forefront of text-to-speech research in Greece for nearly two decades. Its unit-selection text-to-speech system is now among the top quality systems for the Greek language at international level.

The Group has developed TtS engines for the Greek language based on all the major approaches, including:

  • formant rule-based, also combined with computational intelligence approaches (e.g. [Raptis & Carayannis, 1997; Raptis et al, 2001]);
  • diphone synthesis;
  • unit-selection, having developed the first unit-selection TtS system for the Greek language;
  • statistical parametric speech synthesis, performing research on statistical parametric speech synthesis which lead to the first HMM synthesizer for the Greek language ([Karabetsos et al, 2008]) based on the HTS platform.

A set of audio samples from the various systems that the Group has developed are available.

The Group is performing basic and applied research on various aspects of speech synthesis, including speech processing, analysis and modeling. Parametric and expressive speech synthesis are currently among the most active strands of research while research continues in various specific topics such as analysis, spectral estimation and feature extraction. More information on the research outcomes can be found in the publications page.

Focus is also given to the applications of speech synthesis in areas such as speech-enabling of digital content, natural spoken interaction and accessibility.

Further to the above, the Group has worked on transferring its TtS platform to different languages. As a result, a top-quality system has been developed for Bulgarian ([Raptis et al., 2009a; Chalamandaris et al., 2009a]).

At the latest Blizzard Challenge 2010, an international competition aimed to better understand and compare research techniques in building and evaluating corpus-based speech synthesizers on the same data, among the 17 participating systems the system employed by the Group for British English has obtained the 6th position in hub EH1 (the 'rjs' 5-hours database) and shared the 4th position in hub EH2 (the 1-hour 'Roger' database) [Raptis et al., 2010].

The Group has been involved in a number of national and European projects in its fields of interest.

In 2006 innoetics was founded; a spin-off company of ILSP aiming to capitalize on research outcomes in the area of text-to-speech synthesis and exploit them in the context of innovative, state-of-the-art solutions that will offer enhanced multimodal interactivity and will drastically affect the way we communicate, learn and entertain.

To contact the Group use the information in this link.


Powered by Plone, the Open Source Content Management System

This site conforms to the following standards: