Unsupervised Online Adaptation for Speaker Verification over the Telephone

Abstract : This paper presents experiments of unsupervised adaptation for a speaker detection system. The system used is a standard speaker verification system based on cepstral features and Gaus-sian mixture models. Experiments were performed on cellular speech data taken from the NIST 2002 speaker detection evaluation. There was a total of about 30.000 trials involving 330 target speakers and more than 90% of impostor trials. Unsu-pervised adaptation significantly increases the system accuracy, with a reduction of the minimal detection cost function (DCF) from 0.33 for the baseline system to 0.25 with unsupervised on-line adaptation. Two incremental adaptation modes were tested, either by using a fixed decision threshold for adaptation, or by using the a posteriori probability of the true target for weight-ing the adaptation. Both methods provide similar results in the best configurations, but the latter is less sensitive to the actual threshold value.
Document type :
Conference papers
Complete list of metadatas

Cited literature [9 references]  Display  Hide  Download

Contributor : Sylvain Meignier <>
Submitted on : Tuesday, March 28, 2017 - 2:06:16 PM
Last modification on : Saturday, May 4, 2019 - 1:19:36 AM
Long-term archiving on : Thursday, June 29, 2017 - 4:35:29 PM


Publisher files allowed on an open archive


  • HAL Id : hal-01434306, version 1



Claude Barras, Sylvain Meignier, Jean-Luc Gauvain. Unsupervised Online Adaptation for Speaker Verification over the Telephone. The Speaker and Language Recognition Workshop (Odyssey 2004), May 2004, Tolède, Spain. pp.4. ⟨hal-01434306⟩



Record views


Files downloads