Publication Details
A novel estimation of feature-space MLLR for full_covariance models
Povey Daniel
Agarwal Mohit
Akyazi Pinar
Burget Lukáš, doc. Ing., Ph.D. (DCGM)
Feng Kai
Glembek Ondřej, Ing., Ph.D.
Goel Nagendra
Karafiát Martin, Ing., Ph.D. (DCGM)
Rastrow Ariya
Rose Richard
Schwarz Petr, Ing., Ph.D. (DCGM)
Thomas Samuel
Speech recognition, Speaker adaptation, Hidden Markov models, Optimization methods, Linear algebra
The paper is on a novel estimation of feature-space MLLR for full-covariance models. We present a new approach for full-covariance Gaussian models.
In this paper we present a novel approach for estimating featurespace maximum likelihood linear regression (fMLLR) transforms for full-covariance Gaussian models by directly maximizing the likelihood function by repeated line search in the direction of the gradient. We do this in a pre-transformed parameter space such that an approximation to the expected Hessian is proportional to the unit matrix. The proposed algorithm is as efficient or more efficient than standard approaches, and is more flexible because it can naturally be combined with sets of basis transforms and with full covariance and subspace precision and mean (SPAM) models.
@inproceedings{BUT37049,
author="Arnab {Ghoshal} and Daniel {Povey} and Mohit {Agarwal} and Pinar {Akyazi} and Lukáš {Burget} and Kai {Feng} and Ondřej {Glembek} and Nagendra {Goel} and Martin {Karafiát} and Ariya {Rastrow} and Richard {Rose} and Petr {Schwarz} and Samuel {Thomas}",
title="A novel estimation of feature-space MLLR for full_covariance models",
booktitle="Proc. International Conference on Acoustics, Speech, and Signal Processing",
year="2010",
journal="Proc. International Conference on Acoustics, Speech, and Signal Processing",
volume="2010",
number="3",
pages="4310--4313",
publisher="IEEE Signal Processing Society",
address="Dallas",
isbn="978-1-4244-4296-6",
issn="1520-6149",
url="http://www.fit.vutbr.cz/research/groups/speech/publi/2010/ghoshal_icassp2010_4310.pdf"
}