Loading…

Reverberation aware deep learning for environment tolerant microphone array DOA estimation

By transferring the direction of arrival (DOA) estimation into a problem of spatial sample classification, deep neural network (DNN) driven by large training data have been investigated to achieve significant performance enhancement compared to classic DOA estimation strategies do. However, consider...

Full description

Saved in:
Bibliographic Details
Published in:Applied acoustics 2021-12, Vol.184, p.108337, Article 108337
Main Authors: Liu, Yuji, Tong, Feng, Zhong, Shuanglian, Hong, Qingyang, Li, Lin
Format: Article
Language:English
Subjects:
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
cited_by cdi_FETCH-LOGICAL-c312t-cc13d0287c2120c8862e83486ee365f8c8394bc81fce327564b8ba9ee5fcda8c3
cites cdi_FETCH-LOGICAL-c312t-cc13d0287c2120c8862e83486ee365f8c8394bc81fce327564b8ba9ee5fcda8c3
container_end_page
container_issue
container_start_page 108337
container_title Applied acoustics
container_volume 184
creator Liu, Yuji
Tong, Feng
Zhong, Shuanglian
Hong, Qingyang
Li, Lin
description By transferring the direction of arrival (DOA) estimation into a problem of spatial sample classification, deep neural network (DNN) driven by large training data have been investigated to achieve significant performance enhancement compared to classic DOA estimation strategies do. However, consider the highly diverse indoor environments that microphone arrays being used, different reverberation pattern under various environment pose extremely significant limitations to the DOA network model, as its performance is dependent on the match between training data and practical data. In order to improve the robustness of DOA estimation, a reverberation aware network (RAN) algorithm is proposed in this paper. The beam cross-correlation (BCC) is defined as the cross-correlation between the beamforming output and the observed multi-channel signal, which carries an approximation of the room impulse response. Adopting covariance matrix of BCC matrix as the DNN input to enable explicit characterization of reverberation in which the speech was captured, a reverberation aware deep learning network is trained to improve the generalization capability. Finally, simulation and experiment are performed to verify the effectiveness of the proposed method in improving the environment tolerance of DOA estimation, by comparing to the classic traditional DOA estimation algorithm and MLP strategy.
doi_str_mv 10.1016/j.apacoust.2021.108337
format article
fullrecord <record><control><sourceid>elsevier_cross</sourceid><recordid>TN_cdi_crossref_primary_10_1016_j_apacoust_2021_108337</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><els_id>S0003682X2100431X</els_id><sourcerecordid>S0003682X2100431X</sourcerecordid><originalsourceid>FETCH-LOGICAL-c312t-cc13d0287c2120c8862e83486ee365f8c8394bc81fce327564b8ba9ee5fcda8c3</originalsourceid><addsrcrecordid>eNqFkN1KAzEQhYMoWKuvIHmBrfnp7qZ3lqpVKBREoXgTsrOzmtImSxIrfXtTq9dezcxhzuHwEXLN2YgzXt2sR6Y34D9jGgkmeBaVlPUJGXBVi2LC2eqUDBhjsqiUWJ2TixjX-WSiLAfk7Rl3GBoMJlnvqPkyAWmL2NMNmuCse6edDxTdzgbvtugSTX6T3_OytRB8_-EdUhOC2dO75ZRiTHb7E3ZJzjqziXj1O4fk9eH-ZfZYLJbzp9l0UYDkIhUAXLZMqBoEFwyUqgQqOVYVoqzKToGSk3EDineAUtRlNW5UYyaIZQetUSCHpDrm5jYxBux0H3KFsNec6QMhvdZ_hPSBkD4SysbboxFzu53FoCNYdICtDQhJt97-F_ENfI51PA</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Reverberation aware deep learning for environment tolerant microphone array DOA estimation</title><source>Elsevier</source><creator>Liu, Yuji ; Tong, Feng ; Zhong, Shuanglian ; Hong, Qingyang ; Li, Lin</creator><creatorcontrib>Liu, Yuji ; Tong, Feng ; Zhong, Shuanglian ; Hong, Qingyang ; Li, Lin</creatorcontrib><description>By transferring the direction of arrival (DOA) estimation into a problem of spatial sample classification, deep neural network (DNN) driven by large training data have been investigated to achieve significant performance enhancement compared to classic DOA estimation strategies do. However, consider the highly diverse indoor environments that microphone arrays being used, different reverberation pattern under various environment pose extremely significant limitations to the DOA network model, as its performance is dependent on the match between training data and practical data. In order to improve the robustness of DOA estimation, a reverberation aware network (RAN) algorithm is proposed in this paper. The beam cross-correlation (BCC) is defined as the cross-correlation between the beamforming output and the observed multi-channel signal, which carries an approximation of the room impulse response. Adopting covariance matrix of BCC matrix as the DNN input to enable explicit characterization of reverberation in which the speech was captured, a reverberation aware deep learning network is trained to improve the generalization capability. Finally, simulation and experiment are performed to verify the effectiveness of the proposed method in improving the environment tolerance of DOA estimation, by comparing to the classic traditional DOA estimation algorithm and MLP strategy.</description><identifier>ISSN: 0003-682X</identifier><identifier>EISSN: 1872-910X</identifier><identifier>DOI: 10.1016/j.apacoust.2021.108337</identifier><language>eng</language><publisher>Elsevier Ltd</publisher><subject>Beamforming ; Deep learning ; Direction of arrival estimation (DOA) ; Indoor reverberation ; Reverberation aware network (RAN)</subject><ispartof>Applied acoustics, 2021-12, Vol.184, p.108337, Article 108337</ispartof><rights>2021 Elsevier Ltd</rights><lds50>peer_reviewed</lds50><woscitedreferencessubscribed>false</woscitedreferencessubscribed><citedby>FETCH-LOGICAL-c312t-cc13d0287c2120c8862e83486ee365f8c8394bc81fce327564b8ba9ee5fcda8c3</citedby><cites>FETCH-LOGICAL-c312t-cc13d0287c2120c8862e83486ee365f8c8394bc81fce327564b8ba9ee5fcda8c3</cites></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>314,780,784,27924,27925</link.rule.ids></links><search><creatorcontrib>Liu, Yuji</creatorcontrib><creatorcontrib>Tong, Feng</creatorcontrib><creatorcontrib>Zhong, Shuanglian</creatorcontrib><creatorcontrib>Hong, Qingyang</creatorcontrib><creatorcontrib>Li, Lin</creatorcontrib><title>Reverberation aware deep learning for environment tolerant microphone array DOA estimation</title><title>Applied acoustics</title><description>By transferring the direction of arrival (DOA) estimation into a problem of spatial sample classification, deep neural network (DNN) driven by large training data have been investigated to achieve significant performance enhancement compared to classic DOA estimation strategies do. However, consider the highly diverse indoor environments that microphone arrays being used, different reverberation pattern under various environment pose extremely significant limitations to the DOA network model, as its performance is dependent on the match between training data and practical data. In order to improve the robustness of DOA estimation, a reverberation aware network (RAN) algorithm is proposed in this paper. The beam cross-correlation (BCC) is defined as the cross-correlation between the beamforming output and the observed multi-channel signal, which carries an approximation of the room impulse response. Adopting covariance matrix of BCC matrix as the DNN input to enable explicit characterization of reverberation in which the speech was captured, a reverberation aware deep learning network is trained to improve the generalization capability. Finally, simulation and experiment are performed to verify the effectiveness of the proposed method in improving the environment tolerance of DOA estimation, by comparing to the classic traditional DOA estimation algorithm and MLP strategy.</description><subject>Beamforming</subject><subject>Deep learning</subject><subject>Direction of arrival estimation (DOA)</subject><subject>Indoor reverberation</subject><subject>Reverberation aware network (RAN)</subject><issn>0003-682X</issn><issn>1872-910X</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2021</creationdate><recordtype>article</recordtype><recordid>eNqFkN1KAzEQhYMoWKuvIHmBrfnp7qZ3lqpVKBREoXgTsrOzmtImSxIrfXtTq9dezcxhzuHwEXLN2YgzXt2sR6Y34D9jGgkmeBaVlPUJGXBVi2LC2eqUDBhjsqiUWJ2TixjX-WSiLAfk7Rl3GBoMJlnvqPkyAWmL2NMNmuCse6edDxTdzgbvtugSTX6T3_OytRB8_-EdUhOC2dO75ZRiTHb7E3ZJzjqziXj1O4fk9eH-ZfZYLJbzp9l0UYDkIhUAXLZMqBoEFwyUqgQqOVYVoqzKToGSk3EDineAUtRlNW5UYyaIZQetUSCHpDrm5jYxBux0H3KFsNec6QMhvdZ_hPSBkD4SysbboxFzu53FoCNYdICtDQhJt97-F_ENfI51PA</recordid><startdate>20211215</startdate><enddate>20211215</enddate><creator>Liu, Yuji</creator><creator>Tong, Feng</creator><creator>Zhong, Shuanglian</creator><creator>Hong, Qingyang</creator><creator>Li, Lin</creator><general>Elsevier Ltd</general><scope>AAYXX</scope><scope>CITATION</scope></search><sort><creationdate>20211215</creationdate><title>Reverberation aware deep learning for environment tolerant microphone array DOA estimation</title><author>Liu, Yuji ; Tong, Feng ; Zhong, Shuanglian ; Hong, Qingyang ; Li, Lin</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-c312t-cc13d0287c2120c8862e83486ee365f8c8394bc81fce327564b8ba9ee5fcda8c3</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2021</creationdate><topic>Beamforming</topic><topic>Deep learning</topic><topic>Direction of arrival estimation (DOA)</topic><topic>Indoor reverberation</topic><topic>Reverberation aware network (RAN)</topic><toplevel>peer_reviewed</toplevel><toplevel>online_resources</toplevel><creatorcontrib>Liu, Yuji</creatorcontrib><creatorcontrib>Tong, Feng</creatorcontrib><creatorcontrib>Zhong, Shuanglian</creatorcontrib><creatorcontrib>Hong, Qingyang</creatorcontrib><creatorcontrib>Li, Lin</creatorcontrib><collection>CrossRef</collection><jtitle>Applied acoustics</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Liu, Yuji</au><au>Tong, Feng</au><au>Zhong, Shuanglian</au><au>Hong, Qingyang</au><au>Li, Lin</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Reverberation aware deep learning for environment tolerant microphone array DOA estimation</atitle><jtitle>Applied acoustics</jtitle><date>2021-12-15</date><risdate>2021</risdate><volume>184</volume><spage>108337</spage><pages>108337-</pages><artnum>108337</artnum><issn>0003-682X</issn><eissn>1872-910X</eissn><abstract>By transferring the direction of arrival (DOA) estimation into a problem of spatial sample classification, deep neural network (DNN) driven by large training data have been investigated to achieve significant performance enhancement compared to classic DOA estimation strategies do. However, consider the highly diverse indoor environments that microphone arrays being used, different reverberation pattern under various environment pose extremely significant limitations to the DOA network model, as its performance is dependent on the match between training data and practical data. In order to improve the robustness of DOA estimation, a reverberation aware network (RAN) algorithm is proposed in this paper. The beam cross-correlation (BCC) is defined as the cross-correlation between the beamforming output and the observed multi-channel signal, which carries an approximation of the room impulse response. Adopting covariance matrix of BCC matrix as the DNN input to enable explicit characterization of reverberation in which the speech was captured, a reverberation aware deep learning network is trained to improve the generalization capability. Finally, simulation and experiment are performed to verify the effectiveness of the proposed method in improving the environment tolerance of DOA estimation, by comparing to the classic traditional DOA estimation algorithm and MLP strategy.</abstract><pub>Elsevier Ltd</pub><doi>10.1016/j.apacoust.2021.108337</doi></addata></record>
fulltext fulltext
identifier ISSN: 0003-682X
ispartof Applied acoustics, 2021-12, Vol.184, p.108337, Article 108337
issn 0003-682X
1872-910X
language eng
recordid cdi_crossref_primary_10_1016_j_apacoust_2021_108337
source Elsevier
subjects Beamforming
Deep learning
Direction of arrival estimation (DOA)
Indoor reverberation
Reverberation aware network (RAN)
title Reverberation aware deep learning for environment tolerant microphone array DOA estimation
url http://sfxeu10.hosted.exlibrisgroup.com/loughborough?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-08T03%3A26%3A16IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-elsevier_cross&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Reverberation%20aware%20deep%20learning%20for%20environment%20tolerant%20microphone%20array%20DOA%20estimation&rft.jtitle=Applied%20acoustics&rft.au=Liu,%20Yuji&rft.date=2021-12-15&rft.volume=184&rft.spage=108337&rft.pages=108337-&rft.artnum=108337&rft.issn=0003-682X&rft.eissn=1872-910X&rft_id=info:doi/10.1016/j.apacoust.2021.108337&rft_dat=%3Celsevier_cross%3ES0003682X2100431X%3C/elsevier_cross%3E%3Cgrp_id%3Ecdi_FETCH-LOGICAL-c312t-cc13d0287c2120c8862e83486ee365f8c8394bc81fce327564b8ba9ee5fcda8c3%3C/grp_id%3E%3Coa%3E%3C/oa%3E%3Curl%3E%3C/url%3E&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true