MULTICHANNEL NOISE REDUCTION USING DILATED MULTICHANNEL U-NET AND PRE-TRAINED SINGLE-CHANNEL NETWORK

Zhi Wei Tan; Anh H.T. Nguyen; Yuan Liu; Andy W.H. Khong

doi:10.1109/ICASSP43922.2022.9747132

MULTICHANNEL NOISE REDUCTION USING DILATED MULTICHANNEL U-NET AND PRE-TRAINED SINGLE-CHANNEL NETWORK

Zhi Wei Tan, Anh H.T. Nguyen, Yuan Liu, Andy W.H. Khong

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

7 Citations (Scopus)

Abstract

Pre-trained single-channel neural networks have become more prevalent for noise reduction in recent years. However, unlike their multichannel counterparts, these monoaural approaches do not exploit spatial information during the optimization process. Furthermore, while multichannel neural networks exploit spatial information, they are optimized for a specific microphone array configuration; extensive data collection and training are required if a new array configuration is deployed. We propose a transfer learning approach that leverages existing pre-trained single-channel neural networks for the optimization of multichannel neural networks. Simulation results on the CHiME-3 dataset show that the proposed method outperforms the state-of-the-art multichannel neural network and neural beamformer.

Original language	English
Title of host publication	2022 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2022 - Proceedings
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	266-270
Number of pages	5
ISBN (Electronic)	9781665405409
DOIs	https://doi.org/10.1109/ICASSP43922.2022.9747132
Publication status	Published - 2022
Externally published	Yes
Event	47th IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2022 - Virtual, Online, Singapore Duration: May 23 2022 → May 27 2022

Publication series

Name	ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings
Volume	2022-May
ISSN (Print)	1520-6149

Conference

Conference	47th IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2022
Country/Territory	Singapore
City	Virtual, Online
Period	5/23/22 → 5/27/22

Bibliographical note

Publisher Copyright:
© 2022 IEEE

ASJC Scopus Subject Areas

Software
Signal Processing
Electrical and Electronic Engineering

Keywords

data scarcity
deep learning
fine-tuning
Multichannel speech enhancement
transfer learning

Access to Document

10.1109/ICASSP43922.2022.9747132

Cite this

Tan, Z. W., Nguyen, A. H. T., Liu, Y., & Khong, A. W. H. (2022). MULTICHANNEL NOISE REDUCTION USING DILATED MULTICHANNEL U-NET AND PRE-TRAINED SINGLE-CHANNEL NETWORK. In 2022 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2022 - Proceedings (pp. 266-270). (ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings; Vol. 2022-May). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ICASSP43922.2022.9747132

Tan, Zhi Wei ; Nguyen, Anh H.T. ; Liu, Yuan et al. / MULTICHANNEL NOISE REDUCTION USING DILATED MULTICHANNEL U-NET AND PRE-TRAINED SINGLE-CHANNEL NETWORK. 2022 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2022 - Proceedings. Institute of Electrical and Electronics Engineers Inc., 2022. pp. 266-270 (ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings).

@inproceedings{f5fa34e307ce4c9192b1d3261f429ada,

title = "MULTICHANNEL NOISE REDUCTION USING DILATED MULTICHANNEL U-NET AND PRE-TRAINED SINGLE-CHANNEL NETWORK",

abstract = "Pre-trained single-channel neural networks have become more prevalent for noise reduction in recent years. However, unlike their multichannel counterparts, these monoaural approaches do not exploit spatial information during the optimization process. Furthermore, while multichannel neural networks exploit spatial information, they are optimized for a specific microphone array configuration; extensive data collection and training are required if a new array configuration is deployed. We propose a transfer learning approach that leverages existing pre-trained single-channel neural networks for the optimization of multichannel neural networks. Simulation results on the CHiME-3 dataset show that the proposed method outperforms the state-of-the-art multichannel neural network and neural beamformer.",

keywords = "data scarcity, deep learning, fine-tuning, Multichannel speech enhancement, transfer learning",

author = "Tan, {Zhi Wei} and Nguyen, {Anh H.T.} and Yuan Liu and Khong, {Andy W.H.}",

note = "Publisher Copyright: {\textcopyright} 2022 IEEE; 47th IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2022 ; Conference date: 23-05-2022 Through 27-05-2022",

year = "2022",

doi = "10.1109/ICASSP43922.2022.9747132",

language = "English",

series = "ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "266--270",

booktitle = "2022 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2022 - Proceedings",

address = "United States",

}

Tan, ZW, Nguyen, AHT, Liu, Y & Khong, AWH 2022, MULTICHANNEL NOISE REDUCTION USING DILATED MULTICHANNEL U-NET AND PRE-TRAINED SINGLE-CHANNEL NETWORK. in 2022 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2022 - Proceedings. ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings, vol. 2022-May, Institute of Electrical and Electronics Engineers Inc., pp. 266-270, 47th IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2022, Virtual, Online, Singapore, 5/23/22. https://doi.org/10.1109/ICASSP43922.2022.9747132

MULTICHANNEL NOISE REDUCTION USING DILATED MULTICHANNEL U-NET AND PRE-TRAINED SINGLE-CHANNEL NETWORK. / Tan, Zhi Wei; Nguyen, Anh H.T.; Liu, Yuan et al.
2022 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2022 - Proceedings. Institute of Electrical and Electronics Engineers Inc., 2022. p. 266-270 (ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings; Vol. 2022-May).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

TY - GEN

T1 - MULTICHANNEL NOISE REDUCTION USING DILATED MULTICHANNEL U-NET AND PRE-TRAINED SINGLE-CHANNEL NETWORK

AU - Tan, Zhi Wei

AU - Nguyen, Anh H.T.

AU - Liu, Yuan

AU - Khong, Andy W.H.

PY - 2022

Y1 - 2022

N2 - Pre-trained single-channel neural networks have become more prevalent for noise reduction in recent years. However, unlike their multichannel counterparts, these monoaural approaches do not exploit spatial information during the optimization process. Furthermore, while multichannel neural networks exploit spatial information, they are optimized for a specific microphone array configuration; extensive data collection and training are required if a new array configuration is deployed. We propose a transfer learning approach that leverages existing pre-trained single-channel neural networks for the optimization of multichannel neural networks. Simulation results on the CHiME-3 dataset show that the proposed method outperforms the state-of-the-art multichannel neural network and neural beamformer.

AB - Pre-trained single-channel neural networks have become more prevalent for noise reduction in recent years. However, unlike their multichannel counterparts, these monoaural approaches do not exploit spatial information during the optimization process. Furthermore, while multichannel neural networks exploit spatial information, they are optimized for a specific microphone array configuration; extensive data collection and training are required if a new array configuration is deployed. We propose a transfer learning approach that leverages existing pre-trained single-channel neural networks for the optimization of multichannel neural networks. Simulation results on the CHiME-3 dataset show that the proposed method outperforms the state-of-the-art multichannel neural network and neural beamformer.

KW - data scarcity

KW - deep learning

KW - fine-tuning

KW - Multichannel speech enhancement

KW - transfer learning

UR - http://www.scopus.com/inward/record.url?scp=85131236729&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85131236729&partnerID=8YFLogxK

U2 - 10.1109/ICASSP43922.2022.9747132

DO - 10.1109/ICASSP43922.2022.9747132

M3 - Conference contribution

AN - SCOPUS:85131236729

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

SP - 266

EP - 270

BT - 2022 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2022 - Proceedings

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 47th IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2022

Y2 - 23 May 2022 through 27 May 2022

ER -

Tan ZW, Nguyen AHT, Liu Y, Khong AWH. MULTICHANNEL NOISE REDUCTION USING DILATED MULTICHANNEL U-NET AND PRE-TRAINED SINGLE-CHANNEL NETWORK. In 2022 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2022 - Proceedings. Institute of Electrical and Electronics Engineers Inc. 2022. p. 266-270. (ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings). doi: 10.1109/ICASSP43922.2022.9747132

MULTICHANNEL NOISE REDUCTION USING DILATED MULTICHANNEL U-NET AND PRE-TRAINED SINGLE-CHANNEL NETWORK

Abstract

Publication series

Conference

Bibliographical note

ASJC Scopus Subject Areas

Keywords

Access to Document

Other files and links

Fingerprint

Cite this