Publication Details

Speech Enhancement Using End-to-End Speech Recognition Objectives

SUBRAMANIAN Aswin S., WANG Xiaofei, BASKAR Murali K., WATANABE Shinji, TANIGUCHI Toru, TRAN Dung and FUJITA Yuya. Speech Enhancement Using End-to-End Speech Recognition Objectives. In: IEEE Workshop on Applications of Signal Processing to Audio and Acoustics. New Paltz, NY: IEEE Signal Processing Society, 2019, pp. 234-238. ISBN 978-1-7281-1123-0. Available from: https://ieeexplore.ieee.org/document/8937250
Czech title
Zvýrazňování řeči pomocí objektivní funkce end-to-end rozpoznávání řeči
Type
conference paper
Language
english
Authors
Subramanian Aswin S. (JHU)
Wang Xiaofei (JHU)
Baskar Murali K. (DCGM FIT BUT)
Watanabe Shinji, Dr. (JHU)
Taniguchi Toru (YAHOO-JP)
Tran Dung (YAHOO-JP)
Fujita Yuya (YAHOO-JP)
URL
Keywords

speech enhancement, speech recognition, neural dereverberation, neural beamformer, training objectives

Abstract

Speech enhancement systems, which denoise and dereverberate distorted signals, are usually optimized based on signal reconstruction objectives including the maximum likelihood and minimum mean square error. However, emergent end-to-end neural methods enable to optimize the speech enhancement system with more applicationoriented objectives. For example, we can jointly optimize speech enhancement and automatic speech recognition (ASR) only with ASR error minimization criteria. The major contribution of this paper is to investigate how a system optimized based on the ASR objective improves the speech enhancement quality on various signal level metrics in addition to the ASR word error rate (WER) metric. We use a recently developed multichannel end-to-end (ME2E) system, which integrates neural dereverberation, beamforming, and attention-based speech recognition within a single neural network. Additionally, we propose to extend the dereverberation sub network of ME2E by dynamically varying the filter order in linear prediction by using reinforcement learning, and extend the beamforming subnetwork by incorporating the estimation of a speech distortion factor. The experiments reveal how well different signal level metrics correlate with the WER metric, and verify that learning-based speech enhancement can be realized by end-to-end ASR training objectives without using parallel clean and noisy data.

Published
2019
Pages
234-238
Proceedings
IEEE Workshop on Applications of Signal Processing to Audio and Acoustics
Conference
IEEE Workshop on Applications of Signal Processing to Audio and Acoustics (WASPAA), New Paltz, NY, US
ISBN
978-1-7281-1123-0
Publisher
IEEE Signal Processing Society
Place
New Paltz, NY, US
DOI
10.1109/WASPAA.2019.8937250
UT WoS
000527800200048
EID Scopus
2-s2.0-85078046877
BibTeX 
@INPROCEEDINGS{FITPUB12472,
   author = "S. Aswin Subramanian and Xiaofei Wang and K. Murali Baskar and Shinji Watanabe and Toru Taniguchi and Dung Tran and Yuya Fujita",
   title = "Speech Enhancement Using End-to-End Speech Recognition Objectives",
   pages = "234--238",
   booktitle = "IEEE Workshop on Applications of Signal Processing to Audio and Acoustics",
   year = 2019,
   location = "New Paltz, NY, US",
   publisher = "IEEE Signal Processing Society",
   ISBN = "978-1-7281-1123-0",
   doi = "10.1109/WASPAA.2019.8937250",
   language = "english",
   url = "https://www.fit.vut.cz/research/publication/12472"
}
Back to top