Streaming Target-Speaker ASR with Neural Transducer
Although recent advances in deep learning technology have boosted automatic speech recognition (ASR) performance in the single-talker case, it remains difficult to recognize multi-talker speech in which many voices overlap. One conventional approach to tackle this problem is to use a cascade of a sp...
Gespeichert in:
Hauptverfasser: | , , , , |
---|---|
Format: | Artikel |
Sprache: | eng |
Schlagworte: | |
Online-Zugang: | Volltext bestellen |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
container_end_page | |
---|---|
container_issue | |
container_start_page | |
container_title | |
container_volume | |
creator | Moriya, Takafumi Sato, Hiroshi Ochiai, Tsubasa Delcroix, Marc Shinozaki, Takahiro |
description | Although recent advances in deep learning technology have boosted automatic
speech recognition (ASR) performance in the single-talker case, it remains
difficult to recognize multi-talker speech in which many voices overlap. One
conventional approach to tackle this problem is to use a cascade of a speech
separation or target speech extraction front-end with an ASR back-end. However,
the extra computation costs of the front-end module are a critical barrier to
quick response, especially for streaming ASR. In this paper, we propose a
target-speaker ASR (TS-ASR) system that implicitly integrates the target speech
extraction functionality within a streaming end-to-end (E2E) ASR system, i.e.
recurrent neural network-transducer (RNNT). Our system uses a similar idea as
adopted for target speech extraction, but implements it directly at the level
of the encoder of RNNT. This allows TS-ASR to be realized without placing extra
computation costs on the front-end. Note that this study presents two major
differences between prior studies on E2E TS-ASR; we investigate streaming
models and base our study on Conformer models, whereas prior studies used
RNN-based systems and considered only offline processing. We confirm in
experiments that our TS-ASR achieves comparable recognition performance with
conventional cascade systems in the offline setting, while reducing computation
costs and realizing streaming TS-ASR. |
doi_str_mv | 10.48550/arxiv.2209.04175 |
format | Article |
fullrecord | <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2209_04175</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2209_04175</sourcerecordid><originalsourceid>FETCH-LOGICAL-a675-1e91c7e4da7bf733a8f4ecc7e5531cc42254cbb5d7f3b85d075fe58b0f26ad293</originalsourceid><addsrcrecordid>eNotzs1OwkAUhuHZuCDoBbBybqB1_k6nXRKiYkIksd03Z2bOYCMQcigqd68iqy95F18eIWZala4GUA_I38NnaYxqSuW0h4mw7ciEu2G_kR3yhsaiPRB-EMt5-ya_hvFdvtKJcSs7xv0xnSLxrbjJuD3S3XWnont67BbLYrV-flnMVwVWHgpNjY6eXEIfsrcW6-wo_hYAq2N0xoCLIUDy2YYakvKQCeqgsqkwmcZOxf3_7UXdH3jYIZ_7P31_0dsfCW8_TA</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Streaming Target-Speaker ASR with Neural Transducer</title><source>arXiv.org</source><creator>Moriya, Takafumi ; Sato, Hiroshi ; Ochiai, Tsubasa ; Delcroix, Marc ; Shinozaki, Takahiro</creator><creatorcontrib>Moriya, Takafumi ; Sato, Hiroshi ; Ochiai, Tsubasa ; Delcroix, Marc ; Shinozaki, Takahiro</creatorcontrib><description>Although recent advances in deep learning technology have boosted automatic
speech recognition (ASR) performance in the single-talker case, it remains
difficult to recognize multi-talker speech in which many voices overlap. One
conventional approach to tackle this problem is to use a cascade of a speech
separation or target speech extraction front-end with an ASR back-end. However,
the extra computation costs of the front-end module are a critical barrier to
quick response, especially for streaming ASR. In this paper, we propose a
target-speaker ASR (TS-ASR) system that implicitly integrates the target speech
extraction functionality within a streaming end-to-end (E2E) ASR system, i.e.
recurrent neural network-transducer (RNNT). Our system uses a similar idea as
adopted for target speech extraction, but implements it directly at the level
of the encoder of RNNT. This allows TS-ASR to be realized without placing extra
computation costs on the front-end. Note that this study presents two major
differences between prior studies on E2E TS-ASR; we investigate streaming
models and base our study on Conformer models, whereas prior studies used
RNN-based systems and considered only offline processing. We confirm in
experiments that our TS-ASR achieves comparable recognition performance with
conventional cascade systems in the offline setting, while reducing computation
costs and realizing streaming TS-ASR.</description><identifier>DOI: 10.48550/arxiv.2209.04175</identifier><language>eng</language><subject>Computer Science - Sound</subject><creationdate>2022-09</creationdate><rights>http://creativecommons.org/licenses/by-nc-nd/4.0</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,776,881</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/2209.04175$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2209.04175$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Moriya, Takafumi</creatorcontrib><creatorcontrib>Sato, Hiroshi</creatorcontrib><creatorcontrib>Ochiai, Tsubasa</creatorcontrib><creatorcontrib>Delcroix, Marc</creatorcontrib><creatorcontrib>Shinozaki, Takahiro</creatorcontrib><title>Streaming Target-Speaker ASR with Neural Transducer</title><description>Although recent advances in deep learning technology have boosted automatic
speech recognition (ASR) performance in the single-talker case, it remains
difficult to recognize multi-talker speech in which many voices overlap. One
conventional approach to tackle this problem is to use a cascade of a speech
separation or target speech extraction front-end with an ASR back-end. However,
the extra computation costs of the front-end module are a critical barrier to
quick response, especially for streaming ASR. In this paper, we propose a
target-speaker ASR (TS-ASR) system that implicitly integrates the target speech
extraction functionality within a streaming end-to-end (E2E) ASR system, i.e.
recurrent neural network-transducer (RNNT). Our system uses a similar idea as
adopted for target speech extraction, but implements it directly at the level
of the encoder of RNNT. This allows TS-ASR to be realized without placing extra
computation costs on the front-end. Note that this study presents two major
differences between prior studies on E2E TS-ASR; we investigate streaming
models and base our study on Conformer models, whereas prior studies used
RNN-based systems and considered only offline processing. We confirm in
experiments that our TS-ASR achieves comparable recognition performance with
conventional cascade systems in the offline setting, while reducing computation
costs and realizing streaming TS-ASR.</description><subject>Computer Science - Sound</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2022</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotzs1OwkAUhuHZuCDoBbBybqB1_k6nXRKiYkIksd03Z2bOYCMQcigqd68iqy95F18eIWZala4GUA_I38NnaYxqSuW0h4mw7ciEu2G_kR3yhsaiPRB-EMt5-ya_hvFdvtKJcSs7xv0xnSLxrbjJuD3S3XWnont67BbLYrV-flnMVwVWHgpNjY6eXEIfsrcW6-wo_hYAq2N0xoCLIUDy2YYakvKQCeqgsqkwmcZOxf3_7UXdH3jYIZ_7P31_0dsfCW8_TA</recordid><startdate>20220909</startdate><enddate>20220909</enddate><creator>Moriya, Takafumi</creator><creator>Sato, Hiroshi</creator><creator>Ochiai, Tsubasa</creator><creator>Delcroix, Marc</creator><creator>Shinozaki, Takahiro</creator><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20220909</creationdate><title>Streaming Target-Speaker ASR with Neural Transducer</title><author>Moriya, Takafumi ; Sato, Hiroshi ; Ochiai, Tsubasa ; Delcroix, Marc ; Shinozaki, Takahiro</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a675-1e91c7e4da7bf733a8f4ecc7e5531cc42254cbb5d7f3b85d075fe58b0f26ad293</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2022</creationdate><topic>Computer Science - Sound</topic><toplevel>online_resources</toplevel><creatorcontrib>Moriya, Takafumi</creatorcontrib><creatorcontrib>Sato, Hiroshi</creatorcontrib><creatorcontrib>Ochiai, Tsubasa</creatorcontrib><creatorcontrib>Delcroix, Marc</creatorcontrib><creatorcontrib>Shinozaki, Takahiro</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Moriya, Takafumi</au><au>Sato, Hiroshi</au><au>Ochiai, Tsubasa</au><au>Delcroix, Marc</au><au>Shinozaki, Takahiro</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Streaming Target-Speaker ASR with Neural Transducer</atitle><date>2022-09-09</date><risdate>2022</risdate><abstract>Although recent advances in deep learning technology have boosted automatic
speech recognition (ASR) performance in the single-talker case, it remains
difficult to recognize multi-talker speech in which many voices overlap. One
conventional approach to tackle this problem is to use a cascade of a speech
separation or target speech extraction front-end with an ASR back-end. However,
the extra computation costs of the front-end module are a critical barrier to
quick response, especially for streaming ASR. In this paper, we propose a
target-speaker ASR (TS-ASR) system that implicitly integrates the target speech
extraction functionality within a streaming end-to-end (E2E) ASR system, i.e.
recurrent neural network-transducer (RNNT). Our system uses a similar idea as
adopted for target speech extraction, but implements it directly at the level
of the encoder of RNNT. This allows TS-ASR to be realized without placing extra
computation costs on the front-end. Note that this study presents two major
differences between prior studies on E2E TS-ASR; we investigate streaming
models and base our study on Conformer models, whereas prior studies used
RNN-based systems and considered only offline processing. We confirm in
experiments that our TS-ASR achieves comparable recognition performance with
conventional cascade systems in the offline setting, while reducing computation
costs and realizing streaming TS-ASR.</abstract><doi>10.48550/arxiv.2209.04175</doi><oa>free_for_read</oa></addata></record> |
fulltext | fulltext_linktorsrc |
identifier | DOI: 10.48550/arxiv.2209.04175 |
ispartof | |
issn | |
language | eng |
recordid | cdi_arxiv_primary_2209_04175 |
source | arXiv.org |
subjects | Computer Science - Sound |
title | Streaming Target-Speaker ASR with Neural Transducer |
url | https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-02-02T10%3A29%3A17IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-arxiv_GOX&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Streaming%20Target-Speaker%20ASR%20with%20Neural%20Transducer&rft.au=Moriya,%20Takafumi&rft.date=2022-09-09&rft_id=info:doi/10.48550/arxiv.2209.04175&rft_dat=%3Carxiv_GOX%3E2209_04175%3C/arxiv_GOX%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true |