Článek ve sborníku konference

 
Grézl, F., Karafiát, M., Burget, L.: Investigation into bottle-neck features for meeting speech recognition, In: Proc. Interspeech 2009, Brighton, GB, ISCA, 2009, s. 2947-2950, ISBN 978-1-61567-692-7, ISSN 1990-9772
Jazyk publikace:angličtina
Název publikace:Investigation into bottle-neck features for meeting speech recognition
Název (cs):Výzkum bottle-neck parametrů v rozpoznávání řeči z meetingů
Strany:2947-2950
Sborník:Proc. Interspeech 2009
Konference:Interspeech 2009
Místo vydání:Brighton, GB
Rok:2009
ISBN:978-1-61567-692-7
Časopis:Proceedings of Interspeech, č. 9, FR
ISSN:1990-9772
Vydavatel:International Speech Communication Association
URL:http://www.fit.vutbr.cz/research/groups/speech/publi/2009/grezl_is2009.pdf [PDF]
Klíčová slova
Bottle-neck, ANN architecture, features, LVCSR
Anotace
Článek je o výzkumu bottle-neck parametrů v rozpoznávání řeči z meetingů. Architektura bottle-neck je portována do architektury využívající rozděleného kontextu, čímž dosahujeme značné snížení chybovosti.
Abstrakt
This work investigates into recently proposed Bottle-Neck features for ASR. The bottle-neck ANN structure is imported into Split Context architecture gaining significant WER reduction. Further, Universal Context architecture was developed which simplifies the system by using only one universal ANN for all temporal splits. Significant WER reduction can be obtained by applying fMPE on top of our BN features as a technique for discriminative feature extraction and further gain is also obtained by retraining model parameters using MPE criterion. The results are reported on meeting data from RT07 evaluation
BibTeX:
@INPROCEEDINGS{
   author = {František Grézl and Martin Karafiát and Lukáš Burget},
   title = {Investigation into bottle-neck features for meeting speech
	recognition},
   pages = {2947--2950},
   booktitle = {Proc. Interspeech 2009},
   journal = {Proceedings of Interspeech},
   number = {9},
   year = {2009},
   location = {Brighton, GB},
   publisher = {International Speech Communication Association},
   ISBN = {978-1-61567-692-7},
   ISSN = {1990-9772},
   language = {english},
   url = {http://www.fit.vutbr.cz/research/view_pub.php?id=9038}
}