Publication Details
Comparison of Keyword Spotting Approaches for Informal Continuous Speech
Schwarz Petr, Ing., Ph.D. (DCGM)
Burget Lukáš, doc. Ing., Ph.D. (DCGM)
Fapšo Michal, Ing., Ph.D.
Karafiát Martin, Ing., Ph.D. (DCGM)
Černocký Jan, prof. Dr. Ing. (DCGM)
Matějka Pavel, Ing., Ph.D. (DCGM)
comparison, keyword spotting, hidden Markov model, long temporal trajectory, phoneme recognizer
This paper describes several approaches to keyword spotting(KWS) for informal continuous speech. We compare acoustickeyword spotting, spotting in word lattices generated by largevocabulary continuous speech recognition and a hybrid approach makinguse of phoneme lattices generated by a phoneme recognizer. The systems are compared on carefully defined test data extracted from ICSImeeting database. The acoustic and phoneme-lattice based KWS arebased on a phoneme recognizer making use of temporal-pattern (TRAP)feature extraction and posterior estimation using neural nets. We showits superiority over traditional HMM/GMM systems. The advantages anddrawbacks of different approaches are discussed.
@inproceedings{BUT18062,
author="Igor {Szőke} and Petr {Schwarz} and Lukáš {Burget} and Michal {Fapšo} and Martin {Karafiát} and Jan {Černocký} and Pavel {Matějka}",
title="Comparison of Keyword Spotting Approaches for Informal Continuous Speech",
booktitle="Interspeech'2005 - Eurospeech - 9th European Conference on Speech Communication and Technology",
year="2005",
journal="European Conference EUROSPEECH",
pages="633--636",
address="Lisabon",
issn="1018-4074",
url="https://www.fit.vut.cz/research/publication/7886/"
}