A Bayesian approach to the verification problem: Applications to speaker verification

In this paper, we study the general verification problem from a Bayesian vi ewpoint. In the Bayesian approach, the verification decision is made by eva luating Bayes factors against a critical threshold. The calculation of the Bayes factors in turn requires the computation of several Bayesian predicti ve densities. As a case study, we apply the method to speaker verification based on the Gaussian mixture model (GMM). We propose an efficient algorith m to calculate the Bayes factors for the GMM, where the Viterbi approximati on is adopted in the computation of joint Bayesian predictive densities. We evaluate the proposed method for the NIST98 speaker verification evaluatio n data. Experimental results show that new Bayesian approach achieves moder ate improvements over a well-trained baseline system using the conventional likelihood ratio test.