- Topic: Fast Speech in Unit Selection Synthesis
- Who: Donata Moers, University of Bonn/University of Bielefeld
- When: March 23rd, (Wednesday); 11.30A.M
- Where: Seminars room (D5-007)
In this presentation of my current work, I present a new approach to synthesizing fast speech in unit selection synthesis.
After recording two inventories - one at normal and one at fast speech rate articulated as accurately as possible - the natural speech first was perceptually evaluated and then automatically preprocessed. Afterwards, utterances were synthesized from both corpora independently. Since fast rate speech differs from normal rate speech in terms of acoustic characteristics, the concept of multi-phone (phoxsy) units (Breuer and Abresch, 2004) was implemented and used to synthesize speech at both speaking rates again. A perceptual evaluation showed that phoxsy units enhanced the intelligibility especially for fast synthetic speech significantly. To conclude with,the ongoing perception experiments investigating fast speech synthesized from the two different inventories and accelerated to varying speech rates with regard to intelligibility and overall speech quality are presented.
--
Heu rebut aquest missatge perquè esteu subscrit al grup "Veu dels dimecres" de Grups de Google.
Per publicar a aquest grup, envieu un correu electrònic a veu-dels...@googlegroups.com.
Per anul·lar la subscripció a aquest grup, envieu un correu electrònic a veu-dels-dimec...@googlegroups.com.
Per obtenir més opcions, visiteu aquest grup a http://groups.google.com/group/veu-dels-dimecres?hl=ca.
Asunción Moreno Directora asuncio...@upc.edu Jordi Girona 1-3, Edificio D5 08034 Barcelona, Spain Tel: (+34) 93 401 64 37 begin_of_the_skype_highlighting (+34) 93 401 64 37 end_of_the_skype_highlighting |
|