Publication Details
Discriminative Acoustic Language Recognition via Channel-Compensated GMM Statistics
Strasheim Albeert
Hubeika Valiantsina, Ing.
Matějka Pavel, Ing., Ph.D. (DCGM)
Burget Lukáš, doc. Ing., Ph.D. (DCGM)
Glembek Ondřej, Ing., Ph.D.
acoustic language recognition, intersession variability compensation, discriminative training
The paper is on Discriminative Acoustic Language Recognition via Channel-Compensated GMM Statistics. The results are reported on NIST LRE'07.
We propose a novel design for acoustic feature-based automatic spoken language recognizers. Our design is inspired by recent advances in text-independent speaker recognition, where intraclass variability is modeled by factor analysis in Gaussian mixture model (GMM) space. We use approximations to GMMlikelihoods which allow variable-length data sequences to be represented as statistics of fixed size. Our experiments on NIST LRE'07 show that variability-compensation of these statistics can reduce error-rates by a factor of three. Finally, we show that further improvements are possible with discriminative logistic regression training.
@inproceedings{BUT33741,
author="Niko {Brümmer} and Albeert {Strasheim} and Valiantsina {Hubeika} and Pavel {Matějka} and Lukáš {Burget} and Ondřej {Glembek}",
title="Discriminative Acoustic Language Recognition via Channel-Compensated GMM Statistics",
booktitle="Proc. Interspeech 2009",
year="2009",
journal="Proceedings of Interspeech",
number="9",
pages="2187--2190",
publisher="International Speech Communication Association",
address="Brighton",
isbn="978-1-61567-692-7",
issn="1990-9772",
url="http://www.fit.vutbr.cz/research/groups/speech/publi/2009/brummer_is2009.pdf"
}