Conditional pronunciation modeling in speaker detection
Authors:
- D. KlusacekDKD. Klusacek
- J. NavratilJNJ. Navratil
- D.A. ReynoldsDRD.A. Reynolds
- J.P. CampbellJCJ.P. Campbell
- 23 January 2004
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 4, 804
Abstract
In this paper, we present a conditional pronunciation mod- eling method for the speaker detection task that does not rely on acoustic vectors. Aiming at exploiting higher- level information carried by the speech signal, it uses time- aligned streams of phones and phonemes to model a spe- aker's specic pronunciation. Our system uses phonemes drawn from a lexicon of pronunciations of words recognized by an automatic speech recognition system to generate the phoneme stream and an open-loop phone recognizer to gen- erate a phone stream. The phoneme and phone streams are aligned at the frame level and conditional probabilities of a phone, given a phoneme, are estimated using co-occurrence counts. A likelihood detector is then applied to these prob- abilities. Performance is measured using the NIST Extended Data paradigm and the Switchboard-I corpus. Using 8 training conversations for enrollment, a 2.1% equal error rate was achieved. Extensions and alternatives, as well as fusion experiments, are presented and discussed.Keywords
This publication has 7 references indexed in Scilit:
- Combining cross-stream and time dimensions in phonetic speaker recognition2004
- Phonetic speaker recognition using maximum-likelihood binary-decision tree models2004
- The SuperSID project: exploiting high-level information for high-accuracy speaker recognition2004
- Speaker verification through large vocabulary continuous speech recognition2002
- Prosody-based automatic segmentation of speech into sentences and topicsSpeech Communication, 2000
- Speaker Recognition on Single- and Multispeaker DataDigital Signal Processing, 2000
- Speaker Verification Using Adapted Gaussian Mixture ModelsDigital Signal Processing, 2000