A Flemish Voice for the Nextens Text-To-Speech System

Wesley Mattheyses, Lukas Latacz, Yuk On Kong, Werner Verhelst

Research output: Chapter in Book/Report/Conference proceedingConference paper

Abstract

Nextens is an open source text-to-speech system that can be used to convert Dutch text into speech as spoken in The Netherlands. Flemish is the variant of Dutch as spoken in Flanders. These two languages have the same written form, but they sound clearly different. This paper describes how we transformed the Nextens system into a Flemish speaking application. In order to achieve this goal, a high-quality acoustic diphone synthesizer has been developed as the new back-end. This synthesizer is based on a very simple and effective overlap-add technique that can be used to simultaneously solve the problem of waveform concatenation and to perform the necessary prosodic modifications. In addition, some post-lex rules have been adapted to the Flemish speaking style. The resulting Flemish diphone synthesis system has a quality that is comparable to that of a commercial diphone synthesis system.
Original languageEnglish
Title of host publicationFifth Slovenian and First International Language Technologies Conference, Ljubljana, Slovenia
Publication statusPublished - 9 Oct 2006
EventFinds and Results from the Swedish Cyprus Expedition: A Gender Perspective at the Medelhavsmuseet - Stockholm, Sweden
Duration: 21 Sep 200925 Sep 2009

Conference

ConferenceFinds and Results from the Swedish Cyprus Expedition: A Gender Perspective at the Medelhavsmuseet
CountrySweden
CityStockholm
Period21/09/0925/09/09

Keywords

  • Text-to-speech
  • speech synthesis

Fingerprint

Dive into the research topics of 'A Flemish Voice for the Nextens Text-To-Speech System'. Together they form a unique fingerprint.

Cite this