Conference paper

RATH Shakti P., KARAFIÁT Martin, GLEMBEK Ondřej and ČERNOCKÝ Jan. A factorized representation of FMLLR transform based on QR-decomposition. In: Proceedings of Interspeech 2012. Portland, Oregon: International Speech Communication Association, 2012, pp. 1-4. ISBN 978-1-62276-759-5. ISSN 1990-9772. Available from: http://www.isca-speech.org/archive/interspeech_2012/i12_0551.html
Publication language:english
Original title:A factorized representation of FMLLR transform based on QR-decomposition
Title (cs):Faktorovaná reprezentace FMLLR transformací založená na QR dekompozici
Pages:1-4
Proceedings:Proceedings of Interspeech 2012
Conference:Interspeech 2012
Place:Portland, Oregon, US
Year:2012
URL:http://www.isca-speech.org/archive/interspeech_2012/i12_0551.html
ISBN:978-1-62276-759-5
Journal:Proceedings of Interspeech, Vol. 2012, No. 9, FR
ISSN:1990-9772
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 QR-decomposition.
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 QR-decomposition of a square matrix and hence is referred to as QR-FMLLR. The mathematical expressions and steps for maximum likelihood (ML) estimation of the parameters are presented. The ML estimation of QR-FMLLR 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 QR-FMLLR to be comparable to the standard FMLLR. We conjecture that QR-FMLLR 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
	QR-decomposition},
   pages = {1--4},
   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 = {978-1-62276-759-5},
   ISSN = {1990-9772},
   language = {english},
   url = {http://www.fit.vutbr.cz/research/view_pub.php?id=10092}
}

Your IPv4 address: 54.167.44.32
Switch to IPv6 connection

DNSSEC [dnssec]