Publication Details

Improving Acoustic Based Keyword Spotting Using LVCSR Lattices

MOTLÍČEK Petr, VALENTE Fabio and SZŐKE Igor. Improving Acoustic Based Keyword Spotting Using LVCSR Lattices. In: Proc. International Conference on Acoustics, Speech, and Signal Processing 2012. Kyoto: IEEE Signal Processing Society, 2012, pp. 4413-4416. ISBN 978-1-4673-0044-5.
Czech title
Vylepšení akustické detekce klíčových slov pomocí LVCSR svazů
Type
conference paper
Language
english
Authors
Motlíček Petr, doc. Ing., Ph.D. (DCGM FIT BUT)
Valente Fabio (IDIAP)
Szőke Igor, Ing., Ph.D. (DCGM FIT BUT)
URL
Keywords

KeyWord Spotting (KWS), Spoken Term Detection (STD), Confidence Measure (CM)

Abstract

This paper summarizes experimental results achieved with acoustic and LVCSR-KWS systems exploited on conversational audio recordings.

Annotation

This paper investigates detection of English keywords in a conversational scenario using a combination of acoustic and LVCSR based keyword spotting systems. Acoustic KWS systems search predefined words in parameterized spoken data. Corresponding confidences are represented by likelihood ratios given the keyword models and a background model. First, due to the especially high number of false-alarms, the acoustic KWS system is augmented with confidence measures estimated from corresponding LVCSR lattices. Then, various strategies to combine scores estimated by the acoustic and several LVCSR based KWS systems are explored. We show that a linear regression based combination significantly outperforms other (model-based) techniques. Due to that, the relative number of false-alarms of the combined KWS system decreased by more than 50% compared to the acoustic KWS system. Finally, an attention is also paid to the complexities of the KWS systems enabling them to potentially be exploited in real-detection tasks.

Published
2012
Pages
4413-4416
Proceedings
Proc. International Conference on Acoustics, Speech, and Signal Processing 2012
Conference
The 37th International Conference on Acoustics, Speech, and Signal Processing, Kyoto, JP
ISBN
978-1-4673-0044-5
Publisher
IEEE Signal Processing Society
Place
Kyoto, JP
DOI
BibTeX
@INPROCEEDINGS{FITPUB9994,
   author = "Petr Motl\'{i}\v{c}ek and Fabio Valente and Igor Sz\H{o}ke",
   title = "Improving Acoustic Based Keyword Spotting Using LVCSR Lattices",
   pages = "4413--4416",
   booktitle = "Proc. International Conference on Acoustics, Speech, and Signal Processing 2012",
   year = 2012,
   location = "Kyoto, JP",
   publisher = "IEEE Signal Processing Society",
   ISBN = "978-1-4673-0044-5",
   doi = "10.1109/ICASSP.2012.6288898",
   language = "english",
   url = "https://www.fit.vut.cz/research/publication/9994"
}
Back to top