Building Synthetic Voices at L2F

From HLT@INESC-ID

Sérgio Paulo

Date

  • 15:30, October 27, 2006
  • 3th floor meeting room

Speaker

Abstract

We aim to go through the whole set of stages that the creation of a cluster unit selection voice for festival (L2F version) calls for. Those stages consist of:

  • Text prompt selection;
  • Recordings and File edition;
  • Search for OOV within the corpus;
  • Phonetic Segmentation;
  • Creation of utterances' HRGs;
  • Detection of Segmentation Errors;
  • Speech Inventory Pruning;
  • Voice Creation Procedure.