Title: Initialization of fMLLR with sufficient statistics from similar speakers
Other Titles: Inicializace fMLLR statistikami od podobných řečníků
Authors: Zajíc, Zbyněk
Machlica, Lukáš
Müller, Luděk
Issue Date: 2011
Publisher: Springer
Document type: článek
article
URI: http://www.kky.zcu.cz/cs/publications/ZbynekZajic_2011_Initializationof
http://hdl.handle.net/11025/16951
ISBN: 978-3-642-23537-5
ISSN: 0302-9743
Keywords: fMLLR;adaptace;statistiky;rozpoznávání řeči;robustnost;initializace
Keywords in different language: fMLLR;adaptation;statistics;speech recognition;robustness;initialization
Abstract: fMLLR si v porovnání s jinými adaptačními metodami vystačí s malým počtem adaptačních dat. Nicméně extrémně malé množství může vyústit ve špatný odhat adaptační matice. Taková situace se předchází dodáním nějaké informace a-priory. V tomto článku diskutujeme možnost inicializace pomocí statistik od podobných řečníků.
Abstract in different language: One of the most utilized adaptation techniques is the feature Maximum Likelihood Linear Regression (fMLLR). In comparison with other adaptation methods the number of free parameters to be estimated significantly decreases. Thus, the method is well suited for situations with small amount of adaptation data. However, fMLLR still fails in situations with extremely small data sets. Such situations can be solved through proper initialization of fMLLR estimation adding some a-priori information. In this paper a novel approach is proposed solving the problem of fMLLR initialization involving statistics from speakers acoustically close to the speaker to be adapted. Proposed initialization suitably substitutes missing adaptation data with similar data from a training database, fMLLR estimation becomes well-conditioned, and the accuracy of the recognition system increases even in situations with extremely small data sets.
Rights: © Zbyněk Zajíc - Lukáš Machlica - Luděk Müller
Appears in Collections:Články / Articles (KKY)
Články / Articles (NTIS)

Files in This Item:
File Description SizeFormat 
ZbynekZajic_2011_Initializationof.pdfPlný text222,99 kBAdobe PDFView/Open


Please use this identifier to cite or link to this item: http://hdl.handle.net/11025/16951

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.