Conference paperRATH Shakti P., KARAFIÁT Martin, GLEMBEK Ondřej and ČERNOCKÝ Jan. A factorized representation of FMLLR transform based on QRdecomposition. In: Proceedings of Interspeech 2012. Portland, Oregon: International Speech Communication Association, 2012, pp. 14. ISBN 9781622767595. ISSN 19909772. Available from: http://www.iscaspeech.org/archive/interspeech_2012/i12_0551.html  Publication language:  english 

Original title:  A factorized representation of FMLLR transform based on QRdecomposition 

Title (cs):  Faktorovaná reprezentace FMLLR transformací založená na QR dekompozici 

Pages:  14 

Proceedings:  Proceedings of Interspeech 2012 

Conference:  Interspeech 2012 

Place:  Portland, Oregon, US 

Year:  2012 

URL:  http://www.iscaspeech.org/archive/interspeech_2012/i12_0551.html 

ISBN:  9781622767595 

Journal:  Proceedings of Interspeech, Vol. 2012, No. 9, FR 

ISSN:  19909772 

Publisher:  International Speech Communication Association 

URL:  http://www.fit.vutbr.cz/research/groups/speech/publi/2012/rath_interspeech2012_618_pp1_4.pdf [PDF] 

Keywords 

FMLLR, QR Decomposition, Orthogonal Matrix,
Givens Rotation, Upper Triangular Matrix 
Annotation 

This paper describes a new factorized representation of FMLLR transform, which is based on QRdecomposition.

Abstract 

In this paper, we propose a novel representation of the FMLLR
transform. This is different from the standard FMLLR in that
the linear transform (LT) is expressed in a factorized form such
that each of the factors involves only one parameter. The representation
is mainly motivated by QRdecomposition of a square
matrix and hence is referred to as QRFMLLR. The mathematical
expressions and steps for maximum likelihood (ML) estimation
of the parameters are presented. The ML estimation of
QRFMLLR does not require the use of numerical technique,
such as gradient ascent, and it does not involve matrix inversion
and computation of matrix determinant. On an LVCSR task, we
show the performance of QRFMLLR to be comparable to the
standard FMLLR. We conjecture that QRFMLLR is amenable
to speaker adaptation using data that varies from very short to
large and present a brief discussion on how this can be achieved.

BibTeX: 

@INPROCEEDINGS{
author = {P. Shakti Rath and Martin Karafi{\'{a}}t and
Ond{\v{r}}ej Glembek and Jan {\v{C}}ernock{\'{y}}},
title = {A factorized representation of FMLLR transform
based on QRdecomposition},
pages = {14},
booktitle = {Proceedings of Interspeech 2012},
journal = {Proceedings of Interspeech},
volume = 2012,
number = 9,
year = 2012,
location = {Portland, Oregon, US},
publisher = {International Speech Communication Association},
ISBN = {9781622767595},
ISSN = {19909772},
language = {english},
url = {http://www.fit.vutbr.cz/research/view_pub.php?id=10092}
} 
