Postfiltering Using an Adversarial Denoising Autoencoder with Noise-aware Training

Naohiro Tawara, Hikari Tanabe, Tetsunori Kobayashi, Masaru Fujieda, Kazuhiro Katagiri, Takashi Yazu, Tetsuji Ogawa

研究成果: Conference contribution

抄録

An adversarial denoising autoencoder (ADAE) with noise-aware training is proposed and successfully applied to post-filtering for linear noise reduction. The ADAE is effective for attenuating interference sounds, however, it is difficult to learn to handle its various unexpected harmful effects (e.g., various types of noise) using a single network. Legacy speech enhancement was introduced as a pre-processor to make it possible to efficiently train the ADAEs by reducing the unexpected variabilities in the inputs to the ADAEs. Time-frequency masking performed well to suppress the variabilities, however, it induced unpleasant distortion, which is difficult for the ADAE to complement. In this paper, a minimum variance distortionless response (MVDR) beam-former, which can avoid troublesome non-linear distortions, is exploited as a preprocessor, and the MVDR outputs are used as the inputs to the ADAE-based post-filter. In addition, noise-dominant signals derived from the MVDR beamformer can improve the accuracy of the ADAE-based post-filter because the residual noise depends on the original noise signals. Experimental comparisons conducted using multichannel speech enhancement demonstrate that ADAE-based post-filtering yields significant improvements over the MVDR-and ADAE-based speech enhancement systems, and noise-aware training of ADAE works well.

元の言語English
ホスト出版物のタイトル2019 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019 - Proceedings
出版者Institute of Electrical and Electronics Engineers Inc.
ページ3282-3286
ページ数5
ISBN(電子版)9781479981311
DOI
出版物ステータスPublished - 2019 5 1
イベント44th IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019 - Brighton, United Kingdom
継続期間: 2019 5 122019 5 17

出版物シリーズ

名前ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings
2019-May
ISSN(印刷物)1520-6149

Conference

Conference44th IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019
United Kingdom
Brighton
期間19/5/1219/5/17

Fingerprint

Speech enhancement
Nonlinear distortion
Noise abatement
Acoustic waves

ASJC Scopus subject areas

  • Software
  • Signal Processing
  • Electrical and Electronic Engineering

これを引用

Tawara, N., Tanabe, H., Kobayashi, T., Fujieda, M., Katagiri, K., Yazu, T., & Ogawa, T. (2019). Postfiltering Using an Adversarial Denoising Autoencoder with Noise-aware Training. : 2019 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019 - Proceedings (pp. 3282-3286). [8682684] (ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings; 巻数 2019-May). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ICASSP.2019.8682684

Postfiltering Using an Adversarial Denoising Autoencoder with Noise-aware Training. / Tawara, Naohiro; Tanabe, Hikari; Kobayashi, Tetsunori; Fujieda, Masaru; Katagiri, Kazuhiro; Yazu, Takashi; Ogawa, Tetsuji.

2019 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019 - Proceedings. Institute of Electrical and Electronics Engineers Inc., 2019. p. 3282-3286 8682684 (ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings; 巻 2019-May).

研究成果: Conference contribution

Tawara, N, Tanabe, H, Kobayashi, T, Fujieda, M, Katagiri, K, Yazu, T & Ogawa, T 2019, Postfiltering Using an Adversarial Denoising Autoencoder with Noise-aware Training. : 2019 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019 - Proceedings., 8682684, ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings, 巻. 2019-May, Institute of Electrical and Electronics Engineers Inc., pp. 3282-3286, 44th IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019, Brighton, United Kingdom, 19/5/12. https://doi.org/10.1109/ICASSP.2019.8682684
Tawara N, Tanabe H, Kobayashi T, Fujieda M, Katagiri K, Yazu T その他. Postfiltering Using an Adversarial Denoising Autoencoder with Noise-aware Training. : 2019 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019 - Proceedings. Institute of Electrical and Electronics Engineers Inc. 2019. p. 3282-3286. 8682684. (ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings). https://doi.org/10.1109/ICASSP.2019.8682684
Tawara, Naohiro ; Tanabe, Hikari ; Kobayashi, Tetsunori ; Fujieda, Masaru ; Katagiri, Kazuhiro ; Yazu, Takashi ; Ogawa, Tetsuji. / Postfiltering Using an Adversarial Denoising Autoencoder with Noise-aware Training. 2019 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019 - Proceedings. Institute of Electrical and Electronics Engineers Inc., 2019. pp. 3282-3286 (ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings).
@inproceedings{741dd6cc4b084cf69146801f8817bba8,
title = "Postfiltering Using an Adversarial Denoising Autoencoder with Noise-aware Training",
abstract = "An adversarial denoising autoencoder (ADAE) with noise-aware training is proposed and successfully applied to post-filtering for linear noise reduction. The ADAE is effective for attenuating interference sounds, however, it is difficult to learn to handle its various unexpected harmful effects (e.g., various types of noise) using a single network. Legacy speech enhancement was introduced as a pre-processor to make it possible to efficiently train the ADAEs by reducing the unexpected variabilities in the inputs to the ADAEs. Time-frequency masking performed well to suppress the variabilities, however, it induced unpleasant distortion, which is difficult for the ADAE to complement. In this paper, a minimum variance distortionless response (MVDR) beam-former, which can avoid troublesome non-linear distortions, is exploited as a preprocessor, and the MVDR outputs are used as the inputs to the ADAE-based post-filter. In addition, noise-dominant signals derived from the MVDR beamformer can improve the accuracy of the ADAE-based post-filter because the residual noise depends on the original noise signals. Experimental comparisons conducted using multichannel speech enhancement demonstrate that ADAE-based post-filtering yields significant improvements over the MVDR-and ADAE-based speech enhancement systems, and noise-aware training of ADAE works well.",
keywords = "Adversarial denoising autoencoder, minimum variance distortionless response, noise-aware training, speech enhancement",
author = "Naohiro Tawara and Hikari Tanabe and Tetsunori Kobayashi and Masaru Fujieda and Kazuhiro Katagiri and Takashi Yazu and Tetsuji Ogawa",
year = "2019",
month = "5",
day = "1",
doi = "10.1109/ICASSP.2019.8682684",
language = "English",
series = "ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
pages = "3282--3286",
booktitle = "2019 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019 - Proceedings",

}

TY - GEN

T1 - Postfiltering Using an Adversarial Denoising Autoencoder with Noise-aware Training

AU - Tawara, Naohiro

AU - Tanabe, Hikari

AU - Kobayashi, Tetsunori

AU - Fujieda, Masaru

AU - Katagiri, Kazuhiro

AU - Yazu, Takashi

AU - Ogawa, Tetsuji

PY - 2019/5/1

Y1 - 2019/5/1

N2 - An adversarial denoising autoencoder (ADAE) with noise-aware training is proposed and successfully applied to post-filtering for linear noise reduction. The ADAE is effective for attenuating interference sounds, however, it is difficult to learn to handle its various unexpected harmful effects (e.g., various types of noise) using a single network. Legacy speech enhancement was introduced as a pre-processor to make it possible to efficiently train the ADAEs by reducing the unexpected variabilities in the inputs to the ADAEs. Time-frequency masking performed well to suppress the variabilities, however, it induced unpleasant distortion, which is difficult for the ADAE to complement. In this paper, a minimum variance distortionless response (MVDR) beam-former, which can avoid troublesome non-linear distortions, is exploited as a preprocessor, and the MVDR outputs are used as the inputs to the ADAE-based post-filter. In addition, noise-dominant signals derived from the MVDR beamformer can improve the accuracy of the ADAE-based post-filter because the residual noise depends on the original noise signals. Experimental comparisons conducted using multichannel speech enhancement demonstrate that ADAE-based post-filtering yields significant improvements over the MVDR-and ADAE-based speech enhancement systems, and noise-aware training of ADAE works well.

AB - An adversarial denoising autoencoder (ADAE) with noise-aware training is proposed and successfully applied to post-filtering for linear noise reduction. The ADAE is effective for attenuating interference sounds, however, it is difficult to learn to handle its various unexpected harmful effects (e.g., various types of noise) using a single network. Legacy speech enhancement was introduced as a pre-processor to make it possible to efficiently train the ADAEs by reducing the unexpected variabilities in the inputs to the ADAEs. Time-frequency masking performed well to suppress the variabilities, however, it induced unpleasant distortion, which is difficult for the ADAE to complement. In this paper, a minimum variance distortionless response (MVDR) beam-former, which can avoid troublesome non-linear distortions, is exploited as a preprocessor, and the MVDR outputs are used as the inputs to the ADAE-based post-filter. In addition, noise-dominant signals derived from the MVDR beamformer can improve the accuracy of the ADAE-based post-filter because the residual noise depends on the original noise signals. Experimental comparisons conducted using multichannel speech enhancement demonstrate that ADAE-based post-filtering yields significant improvements over the MVDR-and ADAE-based speech enhancement systems, and noise-aware training of ADAE works well.

KW - Adversarial denoising autoencoder

KW - minimum variance distortionless response

KW - noise-aware training

KW - speech enhancement

UR - http://www.scopus.com/inward/record.url?scp=85068975809&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85068975809&partnerID=8YFLogxK

U2 - 10.1109/ICASSP.2019.8682684

DO - 10.1109/ICASSP.2019.8682684

M3 - Conference contribution

AN - SCOPUS:85068975809

T3 - ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings

SP - 3282

EP - 3286

BT - 2019 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019 - Proceedings

PB - Institute of Electrical and Electronics Engineers Inc.

ER -