Convolutional Neural Network for Refinement of Speaker Adaptation Transformation

Result code in IS VaVaI
<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F49777513%3A23520%2F14%3A43922932" target="_blank" >RIV/49777513:23520/14:43922932 - isvavai.cz</a>
Result on the web
<a href="http://download.springer.com/static/pdf/914/chp%253A10.1007%252F978-3-319-11581-8_20.pdf?auth66=1413288171_5b620d005701573765a4641007670c58&ext=.pdf" target="_blank" >http://download.springer.com/static/pdf/914/chp%253A10.1007%252F978-3-319-11581-8_20.pdf?auth66=1413288171_5b620d005701573765a4641007670c58&ext=.pdf</a>
DOI - Digital Object Identifier
<a href="http://dx.doi.org/10.1007/978-3-319-11581-8_20" target="_blank" >10.1007/978-3-319-11581-8_20</a>

Result language
angličtina
Original language name
Convolutional Neural Network for Refinement of Speaker Adaptation Transformation
Original language description
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 ispropagated 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.
Czech name
—
Czech description
—

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)

Publication year
2014
Confidentiality
S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Article name in the collection
Speech and Computer, 16th International Conference, SPECOM 2014, Novi Sad, Serbia, October 5-9, 2014, Proceedings
ISBN
978-3-319-11580-1
ISSN
0302-9743
e-ISSN
—
Number of pages
8
Pages from-to
161-168
Publisher name
Springer
Place of publication
Heidelberg
Event location
Novi Sad, Serbia
Event date
Oct 5, 2014
Type of event by nationality
WRD - Celosvětová akce
UT code for WoS article
—

Similar results(10)