Automatic Phoneme Segmentation with Relaxed Textual Constraints

Pierre Lanchantin, Andrew C. Morris, Xavier Rodet, Christophe Veaux

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

Abstract

Speech synthesis by unit selection requires the segmentation of a large single speaker high quality recording. Automatic speech recognition techniques, e.g. Hidden Markov Models (HMM), can be optimised for maximum segmentation accuracy. This paper presents the results of tuning such a phoneme segmentation system. Firstly, using no text transcription, the design of an HMM phoneme recogniser is optimised subject to a phoneme bigram language model. Optimal performance is obtained with triphone models, 7 states per phoneme and 5 Gaussians per state, reaching 94.4% phoneme recognition accuracy with 95.2% of phoneme boundaries within 70 ms of hand labelled boundaries. Secondly, using the textual information modeled by a multi-pronunciation phonetic graph built according to errors found in the first step, the reported phoneme recognition accuracy increases to 96.8% with 96.1% of phoneme boundaries within 70 ms of hand labelled boundaries. Finally, the results from these two segmentation methods based on different phonetic graphs, the evaluation set, the hand labelling and the test procedures are discussed and possible improvements are proposed.
Original languageEnglish
Title of host publicationProceedings of the International Conference on Language Resources and Evaluation, LREC 2008, 26 May - 1 June 2008, Marrakech, Morocco
Place of PublicationMarrakech, Morocco
Publication statusPublished - 2008
EventInternational Conference on Language Resources and Evaluation LREC 2008 - Marrakech, Morocco
Duration: 26 May 20081 Jun 2008

Conference

ConferenceInternational Conference on Language Resources and Evaluation LREC 2008
Country/TerritoryMorocco
CityMarrakech
Period26/05/081/06/08

Keywords

  • corpus
  • speech recognition and understanding
  • speech synthesis
  • text-to-speech systems

Fingerprint

Dive into the research topics of 'Automatic Phoneme Segmentation with Relaxed Textual Constraints'. Together they form a unique fingerprint.

Cite this