A framework for a GMM-UBM based speaker verification and the need of a large arabic database

This paper presents a framework for the Gaussian mixture models-Universal Background Model (GMM-UBM) system, which has proved to be an effective probabilistic model for speaker verification, and has been widely used in most of state-of-the-art systems. In this work we focus on different feature extraction techniques, and different client model training strategies. An experimental evaluation of this framework is done on the TIMIT database. Finally, we explain the need of a large Arabic database in order to estimate the appropriate universal background models UBMpsilas required for the optimal performance of this kind of parameterization.