Double Branches and Stages Neural Network for Joint Acoustic Echo and Noise Suppression

In this letter, we propose a collaborative neural network framework for acoustic echo cancellation (AEC) tasks, where echo paths and spatial information are efficiently modeled through a two-stage subnetwork cascade to jointly repair the complex spectrum of the target speech. Specifically, we design...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:IEEE signal processing letters 2024, Vol.31, p.2150-2154
Hauptverfasser: Zhu, Taohua, Qian, Guobing
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page 2154
container_issue
container_start_page 2150
container_title IEEE signal processing letters
container_volume 31
creator Zhu, Taohua
Qian, Guobing
description In this letter, we propose a collaborative neural network framework for acoustic echo cancellation (AEC) tasks, where echo paths and spatial information are efficiently modeled through a two-stage subnetwork cascade to jointly repair the complex spectrum of the target speech. Specifically, we design a two-path structure consisting of a real part and an imaginary part as well as an amplitude phase, a lightweight network module to suppress part of the echo and potential noise in the first stage, and a larger network to estimate the complex residuals for phase correction and spectral restoration in the second stage. In order to maximize the performance of the model at each stage, we propose two convolutional modules: an inplace gate convolutional module and a complex squeezed temporal convolutional module (CSTCM). In addition, a cross-domain loss function is designed to improve the generalization capability. Experiments are conducted under various mismatch scenarios, and the results show that the proposed dual-path staging method provides superior performance over other advanced methods.
doi_str_mv 10.1109/LSP.2024.3441492
format Article
fullrecord <record><control><sourceid>proquest_RIE</sourceid><recordid>TN_cdi_crossref_primary_10_1109_LSP_2024_3441492</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><ieee_id>10632590</ieee_id><sourcerecordid>3097922362</sourcerecordid><originalsourceid>FETCH-LOGICAL-c175t-4582772562d33c540f9d4144daa0e92fc4ea5e4db87efe510dc289d0f7723553</originalsourceid><addsrcrecordid>eNpNkEtPAjEUhRujiYjuXbho4nrw9sVMl4j4CkETSFw2pXNHBnGK7UyM_94iLFydszjffRxCLhkMGAN9M52_DjhwORBSMqn5EekxpYqMiyE7Th5yyLSG4pScxbgGgIIVqkfe7ny33CC9DbZxK4zUNiWdt_Y92Rl2wW6StN8-fNDKB_rs66alI-e72NaOTtzK_xEzX0ek8267DRhj7ZtzclLZTcSLg_bJ4n6yGD9m05eHp_FomjmWqzaTquB5ztWQl0I4JaHSZTpfltYCal45iVahLJdFjhUqBqXjhS6hSpBQSvTJ9X7sNvivDmNr1r4LTdpoBOhc8_Q-TynYp1zwMQaszDbUnzb8GAZm155J7Zlde-bQXkKu9kiNiP_iQ8GVBvEL4Atqew</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>3097922362</pqid></control><display><type>article</type><title>Double Branches and Stages Neural Network for Joint Acoustic Echo and Noise Suppression</title><source>IEEE Electronic Library Online</source><creator>Zhu, Taohua ; Qian, Guobing</creator><creatorcontrib>Zhu, Taohua ; Qian, Guobing</creatorcontrib><description>In this letter, we propose a collaborative neural network framework for acoustic echo cancellation (AEC) tasks, where echo paths and spatial information are efficiently modeled through a two-stage subnetwork cascade to jointly repair the complex spectrum of the target speech. Specifically, we design a two-path structure consisting of a real part and an imaginary part as well as an amplitude phase, a lightweight network module to suppress part of the echo and potential noise in the first stage, and a larger network to estimate the complex residuals for phase correction and spectral restoration in the second stage. In order to maximize the performance of the model at each stage, we propose two convolutional modules: an inplace gate convolutional module and a complex squeezed temporal convolutional module (CSTCM). In addition, a cross-domain loss function is designed to improve the generalization capability. Experiments are conducted under various mismatch scenarios, and the results show that the proposed dual-path staging method provides superior performance over other advanced methods.</description><identifier>ISSN: 1070-9908</identifier><identifier>EISSN: 1558-2361</identifier><identifier>DOI: 10.1109/LSP.2024.3441492</identifier><identifier>CODEN: ISPLEM</identifier><language>eng</language><publisher>New York: IEEE</publisher><subject>Acoustic echo cancellation ; Acoustics ; complex domain ; Convolution ; deep learning ; double branches and stages ; Logic gates ; Long short term memory ; Microphones ; Modules ; Neural networks ; Noise reduction ; single-channel ; Spatial data ; Task complexity ; Time-frequency analysis</subject><ispartof>IEEE signal processing letters, 2024, Vol.31, p.2150-2154</ispartof><rights>Copyright The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 2024</rights><lds50>peer_reviewed</lds50><woscitedreferencessubscribed>false</woscitedreferencessubscribed><cites>FETCH-LOGICAL-c175t-4582772562d33c540f9d4144daa0e92fc4ea5e4db87efe510dc289d0f7723553</cites><orcidid>0009-0008-8903-1255 ; 0000-0003-0470-0154</orcidid></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktohtml>$$Uhttps://ieeexplore.ieee.org/document/10632590$$EHTML$$P50$$Gieee$$H</linktohtml><link.rule.ids>314,780,784,796,4024,27923,27924,27925,54758</link.rule.ids><linktorsrc>$$Uhttps://ieeexplore.ieee.org/document/10632590$$EView_record_in_IEEE$$FView_record_in_$$GIEEE</linktorsrc></links><search><creatorcontrib>Zhu, Taohua</creatorcontrib><creatorcontrib>Qian, Guobing</creatorcontrib><title>Double Branches and Stages Neural Network for Joint Acoustic Echo and Noise Suppression</title><title>IEEE signal processing letters</title><addtitle>LSP</addtitle><description>In this letter, we propose a collaborative neural network framework for acoustic echo cancellation (AEC) tasks, where echo paths and spatial information are efficiently modeled through a two-stage subnetwork cascade to jointly repair the complex spectrum of the target speech. Specifically, we design a two-path structure consisting of a real part and an imaginary part as well as an amplitude phase, a lightweight network module to suppress part of the echo and potential noise in the first stage, and a larger network to estimate the complex residuals for phase correction and spectral restoration in the second stage. In order to maximize the performance of the model at each stage, we propose two convolutional modules: an inplace gate convolutional module and a complex squeezed temporal convolutional module (CSTCM). In addition, a cross-domain loss function is designed to improve the generalization capability. Experiments are conducted under various mismatch scenarios, and the results show that the proposed dual-path staging method provides superior performance over other advanced methods.</description><subject>Acoustic echo cancellation</subject><subject>Acoustics</subject><subject>complex domain</subject><subject>Convolution</subject><subject>deep learning</subject><subject>double branches and stages</subject><subject>Logic gates</subject><subject>Long short term memory</subject><subject>Microphones</subject><subject>Modules</subject><subject>Neural networks</subject><subject>Noise reduction</subject><subject>single-channel</subject><subject>Spatial data</subject><subject>Task complexity</subject><subject>Time-frequency analysis</subject><issn>1070-9908</issn><issn>1558-2361</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2024</creationdate><recordtype>article</recordtype><sourceid>RIE</sourceid><recordid>eNpNkEtPAjEUhRujiYjuXbho4nrw9sVMl4j4CkETSFw2pXNHBnGK7UyM_94iLFydszjffRxCLhkMGAN9M52_DjhwORBSMqn5EekxpYqMiyE7Th5yyLSG4pScxbgGgIIVqkfe7ny33CC9DbZxK4zUNiWdt_Y92Rl2wW6StN8-fNDKB_rs66alI-e72NaOTtzK_xEzX0ek8267DRhj7ZtzclLZTcSLg_bJ4n6yGD9m05eHp_FomjmWqzaTquB5ztWQl0I4JaHSZTpfltYCal45iVahLJdFjhUqBqXjhS6hSpBQSvTJ9X7sNvivDmNr1r4LTdpoBOhc8_Q-TynYp1zwMQaszDbUnzb8GAZm155J7Zlde-bQXkKu9kiNiP_iQ8GVBvEL4Atqew</recordid><startdate>2024</startdate><enddate>2024</enddate><creator>Zhu, Taohua</creator><creator>Qian, Guobing</creator><general>IEEE</general><general>The Institute of Electrical and Electronics Engineers, Inc. (IEEE)</general><scope>97E</scope><scope>RIA</scope><scope>RIE</scope><scope>AAYXX</scope><scope>CITATION</scope><scope>7SC</scope><scope>7SP</scope><scope>8FD</scope><scope>JQ2</scope><scope>L7M</scope><scope>L~C</scope><scope>L~D</scope><orcidid>https://orcid.org/0009-0008-8903-1255</orcidid><orcidid>https://orcid.org/0000-0003-0470-0154</orcidid></search><sort><creationdate>2024</creationdate><title>Double Branches and Stages Neural Network for Joint Acoustic Echo and Noise Suppression</title><author>Zhu, Taohua ; Qian, Guobing</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-c175t-4582772562d33c540f9d4144daa0e92fc4ea5e4db87efe510dc289d0f7723553</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2024</creationdate><topic>Acoustic echo cancellation</topic><topic>Acoustics</topic><topic>complex domain</topic><topic>Convolution</topic><topic>deep learning</topic><topic>double branches and stages</topic><topic>Logic gates</topic><topic>Long short term memory</topic><topic>Microphones</topic><topic>Modules</topic><topic>Neural networks</topic><topic>Noise reduction</topic><topic>single-channel</topic><topic>Spatial data</topic><topic>Task complexity</topic><topic>Time-frequency analysis</topic><toplevel>peer_reviewed</toplevel><toplevel>online_resources</toplevel><creatorcontrib>Zhu, Taohua</creatorcontrib><creatorcontrib>Qian, Guobing</creatorcontrib><collection>IEEE All-Society Periodicals Package (ASPP) 2005–Present</collection><collection>IEEE All-Society Periodicals Package (ASPP) 1998–Present</collection><collection>IEEE Electronic Library Online</collection><collection>CrossRef</collection><collection>Computer and Information Systems Abstracts</collection><collection>Electronics &amp; Communications Abstracts</collection><collection>Technology Research Database</collection><collection>ProQuest Computer Science Collection</collection><collection>Advanced Technologies Database with Aerospace</collection><collection>Computer and Information Systems Abstracts – Academic</collection><collection>Computer and Information Systems Abstracts Professional</collection><jtitle>IEEE signal processing letters</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Zhu, Taohua</au><au>Qian, Guobing</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Double Branches and Stages Neural Network for Joint Acoustic Echo and Noise Suppression</atitle><jtitle>IEEE signal processing letters</jtitle><stitle>LSP</stitle><date>2024</date><risdate>2024</risdate><volume>31</volume><spage>2150</spage><epage>2154</epage><pages>2150-2154</pages><issn>1070-9908</issn><eissn>1558-2361</eissn><coden>ISPLEM</coden><abstract>In this letter, we propose a collaborative neural network framework for acoustic echo cancellation (AEC) tasks, where echo paths and spatial information are efficiently modeled through a two-stage subnetwork cascade to jointly repair the complex spectrum of the target speech. Specifically, we design a two-path structure consisting of a real part and an imaginary part as well as an amplitude phase, a lightweight network module to suppress part of the echo and potential noise in the first stage, and a larger network to estimate the complex residuals for phase correction and spectral restoration in the second stage. In order to maximize the performance of the model at each stage, we propose two convolutional modules: an inplace gate convolutional module and a complex squeezed temporal convolutional module (CSTCM). In addition, a cross-domain loss function is designed to improve the generalization capability. Experiments are conducted under various mismatch scenarios, and the results show that the proposed dual-path staging method provides superior performance over other advanced methods.</abstract><cop>New York</cop><pub>IEEE</pub><doi>10.1109/LSP.2024.3441492</doi><tpages>5</tpages><orcidid>https://orcid.org/0009-0008-8903-1255</orcidid><orcidid>https://orcid.org/0000-0003-0470-0154</orcidid></addata></record>
fulltext fulltext_linktorsrc
identifier ISSN: 1070-9908
ispartof IEEE signal processing letters, 2024, Vol.31, p.2150-2154
issn 1070-9908
1558-2361
language eng
recordid cdi_crossref_primary_10_1109_LSP_2024_3441492
source IEEE Electronic Library Online
subjects Acoustic echo cancellation
Acoustics
complex domain
Convolution
deep learning
double branches and stages
Logic gates
Long short term memory
Microphones
Modules
Neural networks
Noise reduction
single-channel
Spatial data
Task complexity
Time-frequency analysis
title Double Branches and Stages Neural Network for Joint Acoustic Echo and Noise Suppression
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-25T06%3A27%3A43IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest_RIE&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Double%20Branches%20and%20Stages%20Neural%20Network%20for%20Joint%20Acoustic%20Echo%20and%20Noise%20Suppression&rft.jtitle=IEEE%20signal%20processing%20letters&rft.au=Zhu,%20Taohua&rft.date=2024&rft.volume=31&rft.spage=2150&rft.epage=2154&rft.pages=2150-2154&rft.issn=1070-9908&rft.eissn=1558-2361&rft.coden=ISPLEM&rft_id=info:doi/10.1109/LSP.2024.3441492&rft_dat=%3Cproquest_RIE%3E3097922362%3C/proquest_RIE%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_pqid=3097922362&rft_id=info:pmid/&rft_ieee_id=10632590&rfr_iscdi=true