Skip to Main content Skip to Navigation
Conference papers

Recognition and Rejection Performance in Wordspotting Systems Using Hidden Markov modeling techniques

Yassine Benayed 1 Dominique Fohr 1 Jean-Paul Haton 1 Gérard Chollet 2
1 PAROLE - Analysis, perception and recognition of speech
INRIA Lorraine, LORIA - Laboratoire Lorrain de Recherche en Informatique et ses Applications
2 LHC - Laboratoire Hubert Curien [Saint Etienne]
Abstract : This paper deals with the problem of acceptance/rejection of recognition hypotheses for continuous speech utterances. Two different techniques are investigated to improve the rejection of out-of-vocabulary (OOV) words. A combined approach is first proposed which uses two garbage models (a trained one and an on-line garbage model). The second method uses the trained garbage model and consists in post-processing the recognizer hypotheses by computing for each of them a confidence measure. Both approaches are evaluated in the context of a stock exchange application through the telephone for French. The parameters of the two approaches are studied to improve recognition accuracy.
Mots-clés : spotting modèles de markov cachés keywords hidden markov models
Document type :
Conference papers
Complete list of metadata
https://hal.inria.fr/inria-00100841
Contributor : Publications Loria <>
Submitted on : Tuesday, September 26, 2006 - 2:52:20 PM
Last modification on : Saturday, March 6, 2021 - 3:08:00 AM

Identifiers

  • HAL Id : inria-00100841, version 1

Collections

INRIA | UNIV-ST-ETIENNE | LAHC | CNRS | LORIA | UNIV-LORRAINE | UDL | IOGS | PARISTECH

Citation

Yassine Benayed, Dominique Fohr, Jean-Paul Haton, Gérard Chollet. Recognition and Rejection Performance in Wordspotting Systems Using Hidden Markov modeling techniques. International Workshop speech and computer - SPECOM'2002, Sep 2002, St-Petersburg, Russia, 4 p. ⟨inria-00100841⟩

Export

BibTeX TEI DC DCterms EndNote

Share

Metrics

Record views

231

 

Contact                    Legal Notice                    Personal Data