Rapid feature space MLLR speaker adaptation with bilinear models

Abstract

In this paper, we propose a novel method for rapid feature space Maximum Likelihood Linear Regression (FMLLR) speaker adaptation based on bilinear models. When the amount of adaptation data is limited, the conventional FMLLR transforms can be easily over-trained and can even degrade the performance. In such cases, usually by introducing structural constraints on the FMLLR transformation, the original FMLLR adaptation method can be modified for rapid adaptation. The objective of our bilinear model is to introduce a prior knowledge analysis on the training speakers based on Singular Vector Decomposition (SVD), and to incorporate it in the decoding process. This can effectively reduce the number of free parameters of FMLLR transformation and achieve performance improvements even with limited adaptation data. The efficiency of the proposed algorithm is demonstrated with experiments on the Mandarin digital dataset and the Mandarin voice search dataset respectively.

Fast on-line adaptation using KSVD based acoustic clustering

“…Usually the Gaussian means are re-estimated in majority of the works reported in literature. A couple of attempts have been made to re-estimate even the Gaussian mixtureweights using the basis interpolation approach [5], [6], [7]. In these approaches, the interpolation weights can either be estimated as a global parameter or a number of Gaussians can be tied depending on some criterion and one set of weights is estimated for each class of Gaussians.…”