INTERSPEECH 2004 - ICSLP 8th International Conference on Spoken Language Processing Jeju Island, Korea October 4-8, 2004

Phoxsy: Multi-Phone Segments for Unit Selection Speech Synthesis

Stefan Breuer, Julia Abresch

University of Bonn, Germany

A multi-phone unit specification for unit selection speech synthesis is introduced and tested with regard to its qualitative aspects by means of a listening experiment. This different concept of unit definition aims to prevent spectral discontinuities at highly critical points of concatenation and to allow for a faster creation of speech corpora, as well as a speed-up of cost calculation and unit selection at run time. The new units called phoxsy have been designed for German, but the concept can be easily extended to other languages and may also serve as a basis for new half-phone-like segments.

Full Paper

Bibliographic reference.  Breuer, Stefan / Abresch, Julia (2004): "Phoxsy: multi-phone segments for unit selection speech synthesis", In INTERSPEECH-2004, 1217-1220.