Notice of Retraction Optimal factor analysis with GMM mixture pre-selection

The problem of session variability in text-independent speaker verification has been tackled actively for a few years. Factor analysis has successfully been applied for this problem. However, efficient implementation of factor analysis is also a challenge. In this paper we propose a novel GMM mixture pre-selection based factor analysis model for speaker verification and show how it successfully implemented using some fast approximations which result in minor degradation in accuracy. We tested the novel algorithm on the NIST SRE 2006 evaluation set. With similarly accuracy as factor analysis, the efficiency is improved about 10 times.