Text-to-Speech Synthesis

Text-to-Speech Synthesis

Category: Technologies

The Institute for Language and Speech Processing (ILSP) / "Athena" Research Center is in the forefront of Text-to-Speech (TTS) 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.

Besides that, ILSP has developed TTS engines for the Greek language based on all the major approaches and technologies, including:

  • Rule-based Formant speech synthesis: ILSP has developed a formant synthesizer also combined with computational intelligence approaches.
  • Diphone speech synthesis: ILSP has developed the Ekfonitis+ product, the fist diphone-based synthesizer for the Greek language.
  • Unit-Selection speech synthesis: ILSP has developed the first unit-selection TTS system for the Greek language, offering top and near-natural speech quality.
  • Statistical Parametric speech synthesis: Recently, ILSP has also initiated research on statistical parametric speech synthesis which lead to the first HMM synthesizer for the Greek language based on the HTS platform.

Further to the above, ILSP has worked on transferring its TTS platform to different languages. As a result, a top-quality unit selection system has been developed for the Bulgarian language.
A set of audio samples from the various systems that ILSP has developed are available here: http://speech.ilsp.gr/synthesis/samples.

In addition, basic and applied research is being performed on various aspects of speech synthesis, including speech processing, analysis and modelling. Parametric and expressive speech synthesis is 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.

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.