Publication Details

Robust heteroscedastic linear discriminant analysis and LCRC posterior features in large vocabulary continuous speech recognition

KARAFIÁT Martin, GRÉZL František, SCHWARZ Petr, BURGET Lukáš and ČERNOCKÝ Jan. Robust heteroscedastic linear discriminant analysis and LCRC posterior features in large vocabulary continuous speech recognition. In: Proc. Fifth Slovenian and First International Language Technologies Conference. Ljubljana, 2006, pp. 1-4.
Czech title
Robustní heteroscedastická linearní diskriminační analýza a LCRC posteriorní parametry v rozpoznávání řeči s velkým slovníkem
Type
conference paper
Language
english
Authors
Karafiát Martin, Ing., Ph.D. (DCGM FIT BUT)
Grézl František, Ing., Ph.D. (DCGM FIT BUT)
Schwarz Petr, Ing., Ph.D. (DCGM FIT BUT)
Burget Lukáš, doc. Ing., Ph.D. (DCGM FIT BUT)
Černocký Jan, prof. Dr. Ing. (DCGM FIT BUT)
URL
Keywords

speech processing, LVCSR, feature extraction, posterior features, discriminative transforms

Abstract

The paper is on robust heteroscedastic linear discriminant analysis and LCRC posterior features in large vocabulary continuous speech recognition

Annotation

This paper deals with feature extraction in speech recognition. Three robust variants of popular HLDA transform are investigated. Influence of adding posterior features to PLP feature stream is studied. The experimental results are obtained on CTS (continuous telephone speech) data. Silence-reduced HLDA and LCRC phoneme-state posterior features together provide more than 4% absolute improvement in word error rate.

Published
2006
Pages
1-4
Proceedings
Proc. Fifth Slovenian and First International Language Technologies Conference
Conference
Fifth Slovenian and First International Language Technologies Conference, Ljubljana, SI
Place
Ljubljana, SI
BibTeX 
@INPROCEEDINGS{FITPUB8225,
   author = "Martin Karafi\'{a}t and Franti\v{s}ek Gr\'{e}zl and Petr Schwarz and Luk\'{a}\v{s} Burget and Jan \v{C}ernock\'{y}",
   title = "Robust heteroscedastic linear discriminant analysis and LCRC posterior features in large vocabulary continuous speech recognition",
   pages = "1--4",
   booktitle = "Proc. Fifth Slovenian and First International Language Technologies Conference",
   year = 2006,
   location = "Ljubljana, SI",
   language = "english",
   url = "https://www.fit.vut.cz/research/publication/8225"
}
Back to top