Detail publikace

Comparison of Keyword Spotting Approaches for Informal Continuous Speech

SZŐKE, I., SCHWARZ, P., MATĚJKA, P., BURGET, L., FAPŠO, M., KARAFIÁT, M., ČERNOCKÝ, J.

Originální název

Comparison of Keyword Spotting Approaches for Informal Continuous Speech

Anglický název

Comparison of Keyword Spotting Approaches for Informal Continuous Speech

Jazyk

en

Originální abstrakt

This paper describes several approaches to keyword spotting (KWS) for informal continuous speech. We compare acoustic keyword spotting, spotting in word lattices generated by large vocabulary continuous speech recognition and a hybrid approach making use of phoneme lattices generated by a phoneme recognizer. The systems are compared on carefully defined test data extracted from ICSI meeting database. The advantages and drawbacks of different approaches are discussed. The acoustic and phoneme-lattice based KWS are based on a phoneme recognizer making use of temporal-pattern (TRAP) feature extraction and posterior estimation using neural nets. We show its superiority over traditional HMM/GMM systems. A posterior probability transformation function is introduced for posterior based acoustic keyword spotting. We also propose a posterior masking algorithm to speed-up acoustic keyword spotting.

Anglický abstrakt

This paper describes several approaches to keyword spotting (KWS) for informal continuous speech. We compare acoustic keyword spotting, spotting in word lattices generated by large vocabulary continuous speech recognition and a hybrid approach making use of phoneme lattices generated by a phoneme recognizer. The systems are compared on carefully defined test data extracted from ICSI meeting database. The advantages and drawbacks of different approaches are discussed. The acoustic and phoneme-lattice based KWS are based on a phoneme recognizer making use of temporal-pattern (TRAP) feature extraction and posterior estimation using neural nets. We show its superiority over traditional HMM/GMM systems. A posterior probability transformation function is introduced for posterior based acoustic keyword spotting. We also propose a posterior masking algorithm to speed-up acoustic keyword spotting.

Dokumenty

BibTex


@inproceedings{BUT18063,
  author="Igor {Szőke} and Petr {Schwarz} and Pavel {Matějka} and Lukáš {Burget} and Michal {Fapšo} and Martin {Karafiát} and Jan {Černocký}",
  title="Comparison of Keyword Spotting Approaches for Informal Continuous Speech",
  annote="This paper describes several approaches to keyword spotting (KWS) for
informal continuous speech. We compare acoustic keyword spotting,
spotting in word lattices generated by large vocabulary continuous
speech recognition and a hybrid approach making use of phoneme lattices
generated by a phoneme recognizer. The systems are compared on
carefully defined test data extracted from ICSI meeting database. The
advantages and drawbacks of different approaches are discussed. The
acoustic and phoneme-lattice based KWS are based on a phoneme
recognizer making use of temporal-pattern (TRAP) feature extraction and
posterior estimation using neural nets. We show its superiority over
traditional HMM/GMM systems. A posterior probability transformation
function is introduced for posterior based acoustic keyword spotting.
We also propose a posterior masking algorithm to speed-up acoustic
keyword spotting.

", booktitle="2nd Joint Workshop on Multimodal Interaction and Related Machine Learning Algorithms", chapter="18063", year="2005", month="september", pages="1", type="conference paper" }