Skip to main content

Research Repository

Advanced Search

Two-Microphone Dereverberation for Automatic Speech Recognition of Polish

Kundegorski, Mikolaj; Jackson, Philip J.B.; Ziółko, Bartosz

Two-Microphone Dereverberation for Automatic Speech Recognition of Polish Thumbnail


Authors

Mikolaj Kundegorski

Philip J.B. Jackson

Bartosz Ziółko



Abstract

Reverberation is a common problem for many speech technologies, such as automatic speech recognition (ASR) systems. This paper investigates the novel combination of precedence, binaural and statistical independence cues for enhancing reverberant speech, prior to ASR, under these adverse acoustical conditions when two microphone signals are available. Results of the enhancement are evaluated in terms of relevant signal measures and accuracy for both English and Polish ASR tasks. These show inconsistencies between the signal and recognition measures, although in recognition the proposed method consistently outperforms all other combinations and the spectral-subtraction baseline.

Citation

Kundegorski, M., Jackson, P. J., & Ziółko, B. (2015). Two-Microphone Dereverberation for Automatic Speech Recognition of Polish. Archives of Acoustics, 39(3), 411-420. https://doi.org/10.2478/aoa-2014-0045

Journal Article Type Article
Acceptance Date Aug 2, 2014
Publication Date Mar 1, 2015
Deposit Date Mar 25, 2015
Publicly Available Date Mar 28, 2024
Journal Archives of Acoustics
Print ISSN 0137-5075
Electronic ISSN 2300-262X
Publisher De Gruyter Open
Peer Reviewed Peer Reviewed
Volume 39
Issue 3
Pages 411-420
DOI https://doi.org/10.2478/aoa-2014-0045
Keywords Speech enhancement, Reverberation, ASR, Polish.

Files

Published Journal Article (6 Mb)
PDF

Publisher Licence URL
http://creativecommons.org/licenses/by-nc-nd/4.0/

Copyright Statement
© 2014 Polish Academy of Sciences & Institute of Fundamental Technological Research (IPPT PAN). This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License. (CC BY-NC-ND 3.0)




You might also like



Downloadable Citations