* Position : PhD, 3 years
* Research Area : speech synthesis, prosody modelling
* Location : Orange Labs, Lannion, France
* Start date: Openings Immediate.

* Summary: 
The emergence of corpus-based technologies allowed major improvements in Text-to-Speech (TTS) during the last decade. Such systems can produce very natural synthetic sentences, almost undistinguishable from natural speech. Synthetic prompts can now replace human recordings in some commercial applications, like IVR services. However their use remains delicate due to the lack of prosody control (intonation, rhythm...). The aim of the project is to provide the user with a support tool for easily specifying the prosody of the synthesized speech.

The work will focus on characterising essential prosodic elements needed for expressive speech synthesis, possibly restricted to a specific application domain. The chosen typology will have to match the prosody of the TTS corpora as accurately as possible, through a relevant set of prosodic primitives. The robustness of the topology is critical for automatic annotation of the databases.
The work will also address ergonomics -how to propose to the user a convenient way to specify prosody- and will be closely related to the signal production techniques -signal processing and/or unit selection.

* Research Lab:
The PhD will be hosted in the Speech Synthesis team at Orange Labs. Orange Labs develop a state-of-the-art corpus-based speech synthesizer (demonstrator available on http://tts.elibel.tm.fr).

* Requirements:
The candidate has a (research) master in Computer Science or Electrical Engineering. The candidate has a strong interest in doing research, excellent writing skills in French or English and good programming skills. Knowledge in speech processing or automatic classification is a plus.

* Contacts:
For more information please contact: 
- Cedric Boidin,  , +33 2 96 05 33 53
- Thierry Moudenc,  , +33 2 96 05 16 59