In this paper we explore the use of Variational Bayesian (VB) learning in unsupervised speaker clustering. VB learning is a relatively new learning technique that has the capacity of doing at the same time parameter learning and model selection. We tested this approach on the NIST 1996 HUB-4 evaluation test for speaker clustering when the speaker number is a priori known and when it has to be estimated. VB shows a higher accuracy in terms of average cluster purity and average speaker purity compared to the Maximum Likelihood solution.
Variational Bayesian speaker clustering
ODYSSEY 2004, The speaker and language recognition workshop, May 31-June 3, 2004, Toledo, Spain
© ISCA. Personal use of this material is permitted. The definitive version of this paper was published in ODYSSEY 2004, The speaker and language recognition workshop, May 31-June 3, 2004, Toledo, Spain and is available at :
PERMALINK : https://www.eurecom.fr/publication/1418