A Monte-Carlo method for score normalization in Automatic Speaker Verification using Kullback-Leibler distances

Abstract

In this paper, we propose a new score normalization technique in Automatic Speaker Verification (ASV): the DNorm. The main advantage of this score normalization is that it does not need any additional speech data nor external speaker population, as opposed to the state-ofthe-art approaches. The D-Norm is based on the use of Kullback-Leibler (KL) distances in an ASV context. In a first step, we estimate the KL distances with a MonteCarlo method and we experimentally show that they are correlated with the verification scores. In a second step, we use this correlation to implement a score normalization procedure, the D-Norm. We analyse its performance and we compare it to that of a conventional normalization, the Z-Norm. The results show that performance of the D-Norm is comparable to that of the Z-Norm. We then conclude about the results we obtain and we discuss the applications of this work.

Speaker verification in score-ageing-quality classification space

“…In this respect, the technique is related to the 'Dnorm' (distance normalization) class of normalizing methods (Ben et al, 2002;Yuan et al, 2009). In these works, the distance between each speaker GMM and a UBM is estimated via approximations of the Kullback-Leibler (KL) divergence.…”

Efficient speaker verification system using speaker model clustering for T and Z normalizations

“…A MonteCarlo based symmetric Kullback-Leibler (KL) distance is used to obtain a set of client and impostor data using client model and UBM respectively. However, results presented in [13] show that Z-norm always outperformed D-norm, particularly at low miss-detection rates.…”

“…As opposed to Z-norm, D-norm does not require any speech data to estimate the normalization parameters [13]. Here, pseudo-impostor data is generated using the UBM.…”

A Novel Approach to Combining Client-Dependent and Confidence Information in Multimodal Biometrics

“…In the terms used in [3,13], Z-Norm [13] is impostor-centric (i.e, normalisation is carried out with respect to the impostor distributions calculated "offline" by using additional data), T-Norm [13] is also impostor-centric (but with respect to a given utterance calculated "online" by using additional cohort impostor models). DNorm [14] is neither client-nor impostor-centric; it is specific to the Gaussian Mixture Model (GMM) architecture and is based on Kullback-Leibler distance between two GMM models. In [2], a client-centric version of Z-Norm was proposed.…”