Skip to content

Publications

Detail of publication

Citation

Zbyněk Zajíc and Jan Zelinka and Jan Vaněk and Luděk Müller : Convolutional Neural Network for Refinement of Speaker Adaptation Transformation . 16th International Conference on Speech and Computer, SPECOM 2014, Lecture Notes in Artificial Intelligence, vol. 8773, p. 161-168, 2014.

Abstract

The aim of this work is to propose a refinement of the shift-MLLR (shift Maximum Likelihood Linear Regression) adaptation of an acoustics model in the case of limited amount of adaptation data, which can lead to ill-conditioned transformations matrices. We try to suppress the influence of badly estimated transformation parameters utilizing the Artificial Neural Network (ANN), especially Convolutional Neural Network (CNN) with bottleneck layer on the end. The badly estimated shift-MLLR transformation is propagated through an ANN (suitably trained beforehand), and the output of the net is used as the new refined transformation. To train the ANN the well and the badly conditioned shift-MLLR transformations are used as outputs and inputs of ANN, respectively. Anglická klíčová slova: ASR, Adaptation, shift-MLLR, ANN, CNN, bottleneck

Detail of publication

Title: Convolutional Neural Network for Refinement of Speaker Adaptation Transformation
Author: Zbyněk Zajíc ; Jan Zelinka ; Jan Vaněk ; Luděk Müller
Language: English
Date of publication: 1 Oct 2014
Year: 2014
Type of publication: Papers in proceedings of reviewed conferences
Title of journal or book: 16th International Conference on Speech and Computer, SPECOM 2014
Series: Lecture Notes in Artificial Intelligence
Číslo vydání: 8773
Page: 161 - 168
DOI: 10.1007/978-3-319-11581-8_20
ISBN: 0302-9743
ISSN: 978-3-319-11580-1
Date: 5 Oct 2014 - 9 Oct 2014
/ 2014-11-12 16:24:46 /

Keywords

ASR, Adaptation, shift-MLLR, ANN, CNN, bottleneck

BibTeX

@INPROCEEDINGS{ZbynekZajic_2014_ConvolutionalNeural,
 author = {Zbyn\v{e}k Zaj\'{i}c and Jan Zelinka and Jan Van\v{e}k and Lud\v{e}k M\"{u}ller},
 title = {Convolutional Neural Network for Refinement of Speaker Adaptation Transformation},
 year = {2014},
 journal = {16th International Conference on Speech and Computer, SPECOM 2014},
 volume = {8773},
 pages = {161-168},
 series = {Lecture Notes in Artificial Intelligence},
 ISBN = {0302-9743},
 ISSN = {978-3-319-11580-1},
 doi = {10.1007/978-3-319-11581-8_20},
 url = {http://www.kky.zcu.cz/en/publications/ZbynekZajic_2014_ConvolutionalNeural},
}