Speech Activity Detection: An Economics Approach

TitleSpeech Activity Detection: An Economics Approach
Publication TypeConference Paper
Year of Publication2013
AuthorsTsai, T.. J., & Morgan N.
Other Numbers3599
Abstract

This paper proposes an approach to frame-level speech ac-tivity detection based on the extended metaphor of an eco-nomics marketplace. As in a real marketplace, the simulatedmarketplace encourages features to specialize. Features thatmight not have impressive average performance across the en-tire data set might nonetheless perform very well on a subsetof the data, and the marketplace capitalizes on this special-ization by consulting the features only when their expertiseis relevant. On an experimental data set, we show that theframework is able to effectively utilize the expertise of a set

Acknowledgment

This work was partially supported by funding provided to ICSI by the U.S. Defense Advanced Research Projects Agency (DARPA) under contract number D10PC20024. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors or originators and do not necessarily reflect the views of DARPA or of the U.S. Government.

URLhttps://www.icsi.berkeley.edu/pubs/speech/speechactivity13.pdf
Bibliographic Notes

Proceedings of the IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2013), Vancouver, Canada

Abbreviated Authors

T. J. Tsai and N. Morgan

ICSI Research Group

Speech

ICSI Publication Type

Article in conference proceedings