Speaker Verification Comparison between GMM and GMM-UBM Under Limited Data Condition

Abstract

This work demonstrates the verification of speakers with the Constraint of Limited data (<15 sec). The existing techniques for speaker verification work well for sufficient data (>1 minutes). Developing techniques for verifying the speakers for limited data condition is a challenging issue. In this paper, a comparision study is made using Gaussian Mixture Model (GMM) and GMM-Universal background model (GMM-UBM) with mel-frequency cepstral coefficients (MFCC) as a feature is given. The NIST-2003 database is used to carry-out the experiments. The experiments are conducted using different amount of training and testing data. The experimental results show that GMM-UBM gives a lower equal error rate (EER) compared to GMM.