Discounted likelihood linear regression for rapid speaker adaptation

Gunawardana, Asela; Byrne, William

doi:10.1006/csla.2000.0151

Cited by 3 publications

(2 citation statements)

References 14 publications

(11 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…and setting the result to zero, one can easily show that its general solution is given by (16) Furthermore, if the kernel matrix is symmetric (which is the case in our paper), the solution can be simplified as (17) The regularization parameter can be determined empirically by cross-validation.…”

Section: Regularizationmentioning

confidence: 97%

“…For instance, in discounted likelihood linear regression (DLLR) [16], part of the MLLR likelihood is discounted and interpolated with the likelihood computed from the desired family of distributions that may explain the adaptation data. It was shown that DLLR gave better performance than diagonal, block-diagonal, or full MLLR in Switchboard with as little as 5 s of adaptation speech.…”

mentioning

confidence: 99%

See 1 more Smart Citation

Maximum Penalized Likelihood Kernel Regression for Fast Adaptation

Mak

Lai

Tsang

et al. 2009

IEEE Trans. Audio Speech Lang. Process.

View full text Add to dashboard Cite

Abstract-This paper proposes a nonlinear generalization of the popular maximum-likelihood linear regression (MLLR) adaptation algorithm using kernel methods. The proposed method, called maximum penalized likelihood kernel regression adaptation (MPLKR), applies kernel regression with appropriate regularization to determine the affine model transform in a kernel-induced high-dimensional feature space. Although this is not the first attempt of applying kernel methods to conventional linear adaptation algorithms, unlike most of other kernelized adaptation methods such as kernel eigenvoice or kernel eigen-MLLR, MPLKR has the advantage that it is a convex optimization and its solution is always guaranteed to be globally optimal. In fact, the adapted Gaussian means can be obtained analytically by simply solving a system of linear equations. From the Bayesian perspective, MPLKR can also be considered as the kernel version of maximum a posteriori linear regression (MAPLR) adaptation. Supervised and unsupervised speaker adaptation using MPLKR were evaluated on the Resource Management and Wall Street Journal 5K tasks, respectively, achieving a word error rate reduction of 23.6% and 15.5% respectively over the speaker-independently model.

show abstract

Section: Regularizationmentioning

confidence: 97%

mentioning

confidence: 99%