Initialization of Adaptation by Sufficient Statistics Using Phonetic Tree
The result's identifiers
Result code in IS VaVaI
<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F49777513%3A23520%2F12%3A43915992" target="_blank" >RIV/49777513:23520/12:43915992 - isvavai.cz</a>
Result on the web
—
DOI - Digital Object Identifier
—
Alternative languages
Result language
angličtina
Original language name
Initialization of Adaptation by Sufficient Statistics Using Phonetic Tree
Original language description
In this work we deal with the problem of small amount of data when estimating a feature transformation for the speaker adaptation of an acoustic model. Our goal is to compensate for the lack of adaptation data by a proper initialization of transformationmatrices. Methods used in such situations are described, they are based on collecting additional accumulated statistics from nearest speakers. The proposed initialization approach is based on accumulated statistics too, but it incorporates also phoneticinformation when selecting the ?nearest? statistics. Initialization methods compensating for the absence of actual speaker?s data are tested on telephone recordings with different amounts of adaptation data. In worst situation with extremely small amount of adaptation data relative improvement of 5% is obtained.
Czech name
—
Czech description
—
Classification
Type
D - Article in proceedings
CEP classification
JD - Use of computers, robotics and its application
OECD FORD branch
—
Result continuities
Project
<a href="/en/project/DF12P01OVV022" target="_blank" >DF12P01OVV022: ASR- and MT-based Access to a Large Archive of Cultural Heritage (AMALACH)</a><br>
Continuities
P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)
Others
Publication year
2012
Confidentiality
S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů
Data specific for result type
Article name in the collection
Proceedings 2012 IEEE 11th International Conference on Signal Processing
ISBN
978-1-4673-2194-5
ISSN
—
e-ISSN
—
Number of pages
4
Pages from-to
503-506
Publisher name
IEEE Press
Place of publication
Beijing
Event location
Beijing
Event date
Oct 21, 2012
Type of event by nationality
WRD - Celosvětová akce
UT code for WoS article
—