Personalized speech enhancement combining band-split RNN and speaker attentive module

Personalized speech enhancement combining band-split RNN and speaker attentive module

Target speaker information can be utilized in speech enhancement (SE) models to more effectively extract the desired speech. Previous works introduce the speaker embedding into speech enhancement models by means of concatenation or affine transformation. In this paper, we propose a speaker attentive...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:arXiv.org 2023-03
Hauptverfasser: Le, Xiaohuai, Chen, Li, He, Chao, Guo, Yiqing, Chen, Cheng, Xia, Xianjun, Lu, Jing
Format: Artikel
Sprache:eng
Schlagworte:
Affine transformations
Customization
Embedding
Modules
Signal processing
Speech processing
Speech recognition
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page
container_issue
container_start_page
container_title arXiv.org
container_volume
creator Le, Xiaohuai
Chen, Li
He, Chao
Guo, Yiqing
Chen, Cheng
Xia, Xianjun
Lu, Jing
description Target speaker information can be utilized in speech enhancement (SE) models to more effectively extract the desired speech. Previous works introduce the speaker embedding into speech enhancement models by means of concatenation or affine transformation. In this paper, we propose a speaker attentive module to calculate the attention scores between the speaker embedding and the intermediate features, which are used to rescale the features. By merging this module in the state-of-the-art SE model, we construct the personalized SE model for ICASSP Signal Processing Grand Challenge: DNS Challenge 5 (2023). Our system achieves a final score of 0.529 on the blind test set of track1 and 0.549 on track2.
format Article
fullrecord <record><control><sourceid>proquest</sourceid><recordid>TN_cdi_proquest_journals_2778491991</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2778491991</sourcerecordid><originalsourceid>FETCH-proquest_journals_27784919913</originalsourceid><addsrcrecordid>eNqNzcEKgkAUheEhCJLyHS60FnTU1HUUrSSi1jHqLcfGGZs7tujpM-gBWh1--ODMmMfjOAryhPMF84m6MAz5JuNpGnvsckRLRgsl39gADYh1C6hboWvsUTuoTV9JLfUdKqGbgAYlHZzKEqb6evFAC8K5ycoXQm-aUeGKzW9CEfq_XbL1fnfeHoLBmueI5K6dGe30SleeZXlSREURxf-pDw49QWQ</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2778491991</pqid></control><display><type>article</type><title>Personalized speech enhancement combining band-split RNN and speaker attentive module</title><source>Free E- Journals</source><creator>Le, Xiaohuai ; Chen, Li ; He, Chao ; Guo, Yiqing ; Chen, Cheng ; Xia, Xianjun ; Lu, Jing</creator><creatorcontrib>Le, Xiaohuai ; Chen, Li ; He, Chao ; Guo, Yiqing ; Chen, Cheng ; Xia, Xianjun ; Lu, Jing</creatorcontrib><description>Target speaker information can be utilized in speech enhancement (SE) models to more effectively extract the desired speech. Previous works introduce the speaker embedding into speech enhancement models by means of concatenation or affine transformation. In this paper, we propose a speaker attentive module to calculate the attention scores between the speaker embedding and the intermediate features, which are used to rescale the features. By merging this module in the state-of-the-art SE model, we construct the personalized SE model for ICASSP Signal Processing Grand Challenge: DNS Challenge 5 (2023). Our system achieves a final score of 0.529 on the blind test set of track1 and 0.549 on track2.</description><identifier>EISSN: 2331-8422</identifier><language>eng</language><publisher>Ithaca: Cornell University Library, arXiv.org</publisher><subject>Affine transformations ; Customization ; Embedding ; Modules ; Signal processing ; Speech processing ; Speech recognition</subject><ispartof>arXiv.org, 2023-03</ispartof><rights>2023. This work is published under http://creativecommons.org/licenses/by/4.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>780,784</link.rule.ids></links><search><creatorcontrib>Le, Xiaohuai</creatorcontrib><creatorcontrib>Chen, Li</creatorcontrib><creatorcontrib>He, Chao</creatorcontrib><creatorcontrib>Guo, Yiqing</creatorcontrib><creatorcontrib>Chen, Cheng</creatorcontrib><creatorcontrib>Xia, Xianjun</creatorcontrib><creatorcontrib>Lu, Jing</creatorcontrib><title>Personalized speech enhancement combining band-split RNN and speaker attentive module</title><title>arXiv.org</title><description>Target speaker information can be utilized in speech enhancement (SE) models to more effectively extract the desired speech. Previous works introduce the speaker embedding into speech enhancement models by means of concatenation or affine transformation. In this paper, we propose a speaker attentive module to calculate the attention scores between the speaker embedding and the intermediate features, which are used to rescale the features. By merging this module in the state-of-the-art SE model, we construct the personalized SE model for ICASSP Signal Processing Grand Challenge: DNS Challenge 5 (2023). Our system achieves a final score of 0.529 on the blind test set of track1 and 0.549 on track2.</description><subject>Affine transformations</subject><subject>Customization</subject><subject>Embedding</subject><subject>Modules</subject><subject>Signal processing</subject><subject>Speech processing</subject><subject>Speech recognition</subject><issn>2331-8422</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2023</creationdate><recordtype>article</recordtype><sourceid>ABUWG</sourceid><sourceid>AFKRA</sourceid><sourceid>AZQEC</sourceid><sourceid>BENPR</sourceid><sourceid>CCPQU</sourceid><sourceid>DWQXO</sourceid><recordid>eNqNzcEKgkAUheEhCJLyHS60FnTU1HUUrSSi1jHqLcfGGZs7tujpM-gBWh1--ODMmMfjOAryhPMF84m6MAz5JuNpGnvsckRLRgsl39gADYh1C6hboWvsUTuoTV9JLfUdKqGbgAYlHZzKEqb6evFAC8K5ycoXQm-aUeGKzW9CEfq_XbL1fnfeHoLBmueI5K6dGe30SleeZXlSREURxf-pDw49QWQ</recordid><startdate>20230316</startdate><enddate>20230316</enddate><creator>Le, Xiaohuai</creator><creator>Chen, Li</creator><creator>He, Chao</creator><creator>Guo, Yiqing</creator><creator>Chen, Cheng</creator><creator>Xia, Xianjun</creator><creator>Lu, Jing</creator><general>Cornell University Library, arXiv.org</general><scope>8FE</scope><scope>8FG</scope><scope>ABJCF</scope><scope>ABUWG</scope><scope>AFKRA</scope><scope>AZQEC</scope><scope>BENPR</scope><scope>BGLVJ</scope><scope>CCPQU</scope><scope>DWQXO</scope><scope>HCIFZ</scope><scope>L6V</scope><scope>M7S</scope><scope>PIMPY</scope><scope>PQEST</scope><scope>PQQKQ</scope><scope>PQUKI</scope><scope>PRINS</scope><scope>PTHSS</scope></search><sort><creationdate>20230316</creationdate><title>Personalized speech enhancement combining band-split RNN and speaker attentive module</title><author>Le, Xiaohuai ; Chen, Li ; He, Chao ; Guo, Yiqing ; Chen, Cheng ; Xia, Xianjun ; Lu, Jing</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-proquest_journals_27784919913</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2023</creationdate><topic>Affine transformations</topic><topic>Customization</topic><topic>Embedding</topic><topic>Modules</topic><topic>Signal processing</topic><topic>Speech processing</topic><topic>Speech recognition</topic><toplevel>online_resources</toplevel><creatorcontrib>Le, Xiaohuai</creatorcontrib><creatorcontrib>Chen, Li</creatorcontrib><creatorcontrib>He, Chao</creatorcontrib><creatorcontrib>Guo, Yiqing</creatorcontrib><creatorcontrib>Chen, Cheng</creatorcontrib><creatorcontrib>Xia, Xianjun</creatorcontrib><creatorcontrib>Lu, Jing</creatorcontrib><collection>ProQuest SciTech Collection</collection><collection>ProQuest Technology Collection</collection><collection>Materials Science &amp; Engineering Collection</collection><collection>ProQuest Central (Alumni Edition)</collection><collection>ProQuest Central UK/Ireland</collection><collection>ProQuest Central Essentials</collection><collection>ProQuest Central</collection><collection>Technology Collection</collection><collection>ProQuest One Community College</collection><collection>ProQuest Central Korea</collection><collection>SciTech Premium Collection</collection><collection>ProQuest Engineering Collection</collection><collection>Engineering Database</collection><collection>Access via ProQuest (Open Access)</collection><collection>ProQuest One Academic Eastern Edition (DO NOT USE)</collection><collection>ProQuest One Academic</collection><collection>ProQuest One Academic UKI Edition</collection><collection>ProQuest Central China</collection><collection>Engineering Collection</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Le, Xiaohuai</au><au>Chen, Li</au><au>He, Chao</au><au>Guo, Yiqing</au><au>Chen, Cheng</au><au>Xia, Xianjun</au><au>Lu, Jing</au><format>book</format><genre>document</genre><ristype>GEN</ristype><atitle>Personalized speech enhancement combining band-split RNN and speaker attentive module</atitle><jtitle>arXiv.org</jtitle><date>2023-03-16</date><risdate>2023</risdate><eissn>2331-8422</eissn><abstract>Target speaker information can be utilized in speech enhancement (SE) models to more effectively extract the desired speech. Previous works introduce the speaker embedding into speech enhancement models by means of concatenation or affine transformation. In this paper, we propose a speaker attentive module to calculate the attention scores between the speaker embedding and the intermediate features, which are used to rescale the features. By merging this module in the state-of-the-art SE model, we construct the personalized SE model for ICASSP Signal Processing Grand Challenge: DNS Challenge 5 (2023). Our system achieves a final score of 0.529 on the blind test set of track1 and 0.549 on track2.</abstract><cop>Ithaca</cop><pub>Cornell University Library, arXiv.org</pub><oa>free_for_read</oa></addata></record>
fulltext fulltext
identifier EISSN: 2331-8422
ispartof arXiv.org, 2023-03
issn 2331-8422
language eng
recordid cdi_proquest_journals_2778491991
source Free E- Journals
subjects Affine transformations
Customization
Embedding
Modules
Signal processing
Speech processing
Speech recognition
title Personalized speech enhancement combining band-split RNN and speaker attentive module
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-21T07%3A56%3A49IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest&rft_val_fmt=info:ofi/fmt:kev:mtx:book&rft.genre=document&rft.atitle=Personalized%20speech%20enhancement%20combining%20band-split%20RNN%20and%20speaker%20attentive%20module&rft.jtitle=arXiv.org&rft.au=Le,%20Xiaohuai&rft.date=2023-03-16&rft.eissn=2331-8422&rft_id=info:doi/&rft_dat=%3Cproquest%3E2778491991%3C/proquest%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_pqid=2778491991&rft_id=info:pmid/&rfr_iscdi=true