Total Variability Subspace Adaptation Based Speaker Recognition
Zhi Li · Acta Automatica Sinica · 2014
In text-independent speaker recognition, the identity vector(i-vector) based modeling method has recently been proved to be the most popular and efficient method. It is a key problem to estimate the total variability subspace T efficiently and accurately. In this paper, two adaptation algorithms are proposed in order to improve the performance of the i-vector base system in practical environments. Experiments on the 2008 core speaker recognition evaluation dataset of American NIST and Technology and the self-collected speaker recognition evaluation dataset demonstrate that using the proposed adaptation algorithms to adapt to the total variability subspace T from either the test dataset or the developing dataset is effective for improving the performance. In addition, the combination of the two adaptation algorithms can achieve almost the best performance using the developing dataset rather than the test dataset.