Semi-Supervised Speaker Adaptation for End-to-End Speech Synthesis with Pretrained Models
| العنوان: | Semi-Supervised Speaker Adaptation for End-to-End Speech Synthesis with Pretrained Models |
|---|---|
| المؤلفون: | Sunao Hara, Katsuki Inoue, Shinji Watanabe, Ryuichi Yamamoto, Masanobu Abe, Tomoki Hayashi |
| المصدر: | ICASSP |
| بيانات النشر: | IEEE, 2020. |
| سنة النشر: | 2020 |
| مصطلحات موضوعية: | Similarity (geometry), Computer science, Speech recognition, Speech synthesis, 010501 environmental sciences, computer.software_genre, 01 natural sciences, Pipeline (software), 030507 speech-language pathology & audiology, 03 medical and health sciences, End-to-end principle, Transcription (linguistics), 0305 other medical science, computer, 0105 earth and related environmental sciences, Speaker adaptation |
| الوصف: | Recently, end-to-end text-to-speech (TTS) models have achieved a remarkable performance, however, requiring a large amount of paired text and speech data for training. On the other hand, we can easily collect unpaired dozen minutes of speech recordings for a target speaker without corresponding text data. To make use of such accessible data, the proposed method leverages the recent great success of state-of-the-art end-to-end automatic speech recognition (ASR) systems and obtains corresponding transcriptions from pretrained ASR models. Although these models could only provide text output instead of intermediate linguistic features like phonemes, end-to-end TTS can be well trained with such raw text data directly. Thus, the proposed method can greatly simplify a speaker adaptation pipeline by consistently employing end-to-end ASR/TTS ecosystems. The experimental results show that our proposed method achieved comparable performance to a paired data adaptation method in terms of subjective speaker similarity and objective cepstral distance measures. |
| DOI: | 10.1109/icassp40776.2020.9053371 |
| URL الوصول: | https://explore.openaire.eu/search/publication?articleId=doi_________::858c159b6af41c659715109383e20628 https://doi.org/10.1109/icassp40776.2020.9053371 |
| Rights: | CLOSED |
| رقم الانضمام: | edsair.doi...........858c159b6af41c659715109383e20628 |
| قاعدة البيانات: | OpenAIRE |
| ResultId |
1 |
|---|---|
| Header |
edsair OpenAIRE edsair.doi...........858c159b6af41c659715109383e20628 845 3 unknown 844.669555664063 |
| PLink |
https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&scope=site&db=edsair&AN=edsair.doi...........858c159b6af41c659715109383e20628&custid=s6537998&authtype=sso |
| FullText |
Array
(
[Availability] => 0
)
Array ( [0] => Array ( [Url] => https://explore.openaire.eu/search/publication?articleId=doi_________::858c159b6af41c659715109383e20628# [Name] => EDS - OpenAIRE [Category] => fullText [Text] => View record in OpenAIRE [MouseOverText] => View record in OpenAIRE ) ) |
| Items |
Array
(
[Name] => Title
[Label] => Title
[Group] => Ti
[Data] => Semi-Supervised Speaker Adaptation for End-to-End Speech Synthesis with Pretrained Models
)
Array ( [Name] => Author [Label] => Authors [Group] => Au [Data] => <searchLink fieldCode="AR" term="%22Sunao+Hara%22">Sunao Hara</searchLink><br /><searchLink fieldCode="AR" term="%22Katsuki+Inoue%22">Katsuki Inoue</searchLink><br /><searchLink fieldCode="AR" term="%22Shinji+Watanabe%22">Shinji Watanabe</searchLink><br /><searchLink fieldCode="AR" term="%22Ryuichi+Yamamoto%22">Ryuichi Yamamoto</searchLink><br /><searchLink fieldCode="AR" term="%22Masanobu+Abe%22">Masanobu Abe</searchLink><br /><searchLink fieldCode="AR" term="%22Tomoki+Hayashi%22">Tomoki Hayashi</searchLink> ) Array ( [Name] => TitleSource [Label] => Source [Group] => Src [Data] => ICASSP ) Array ( [Name] => Publisher [Label] => Publisher Information [Group] => PubInfo [Data] => IEEE, 2020. ) Array ( [Name] => DatePubCY [Label] => Publication Year [Group] => Date [Data] => 2020 ) Array ( [Name] => Subject [Label] => Subject Terms [Group] => Su [Data] => <searchLink fieldCode="DE" term="%22Similarity+%28geometry%29%22">Similarity (geometry)</searchLink><br /><searchLink fieldCode="DE" term="%22Computer+science%22">Computer science</searchLink><br /><searchLink fieldCode="DE" term="%22Speech+recognition%22">Speech recognition</searchLink><br /><searchLink fieldCode="DE" term="%22Speech+synthesis%22">Speech synthesis</searchLink><br /><searchLink fieldCode="DE" term="%22010501+environmental+sciences%22">010501 environmental sciences</searchLink><br /><searchLink fieldCode="DE" term="%22computer%2Esoftware%5Fgenre%22">computer.software_genre</searchLink><br /><searchLink fieldCode="DE" term="%2201+natural+sciences%22">01 natural sciences</searchLink><br /><searchLink fieldCode="DE" term="%22Pipeline+%28software%29%22">Pipeline (software)</searchLink><br /><searchLink fieldCode="DE" term="%22030507+speech-language+pathology+%26+audiology%22">030507 speech-language pathology & audiology</searchLink><br /><searchLink fieldCode="DE" term="%2203+medical+and+health+sciences%22">03 medical and health sciences</searchLink><br /><searchLink fieldCode="DE" term="%22End-to-end+principle%22">End-to-end principle</searchLink><br /><searchLink fieldCode="DE" term="%22Transcription+%28linguistics%29%22">Transcription (linguistics)</searchLink><br /><searchLink fieldCode="DE" term="%220305+other+medical+science%22">0305 other medical science</searchLink><br /><searchLink fieldCode="DE" term="%22computer%22">computer</searchLink><br /><searchLink fieldCode="DE" term="%220105+earth+and+related+environmental+sciences%22">0105 earth and related environmental sciences</searchLink><br /><searchLink fieldCode="DE" term="%22Speaker+adaptation%22">Speaker adaptation</searchLink> ) Array ( [Name] => Abstract [Label] => Description [Group] => Ab [Data] => Recently, end-to-end text-to-speech (TTS) models have achieved a remarkable performance, however, requiring a large amount of paired text and speech data for training. On the other hand, we can easily collect unpaired dozen minutes of speech recordings for a target speaker without corresponding text data. To make use of such accessible data, the proposed method leverages the recent great success of state-of-the-art end-to-end automatic speech recognition (ASR) systems and obtains corresponding transcriptions from pretrained ASR models. Although these models could only provide text output instead of intermediate linguistic features like phonemes, end-to-end TTS can be well trained with such raw text data directly. Thus, the proposed method can greatly simplify a speaker adaptation pipeline by consistently employing end-to-end ASR/TTS ecosystems. The experimental results show that our proposed method achieved comparable performance to a paired data adaptation method in terms of subjective speaker similarity and objective cepstral distance measures. ) Array ( [Name] => DOI [Label] => DOI [Group] => ID [Data] => 10.1109/icassp40776.2020.9053371 ) Array ( [Name] => URL [Label] => Access URL [Group] => URL [Data] => <link linkTarget="URL" linkTerm="https://explore.openaire.eu/search/publication?articleId=doi_________::858c159b6af41c659715109383e20628" linkWindow="_blank">https://explore.openaire.eu/search/publication?articleId=doi_________::858c159b6af41c659715109383e20628</link><br /><link linkTarget="URL" linkTerm="https://doi.org/10.1109/icassp40776.2020.9053371" linkWindow="_blank">https://doi.org/10.1109/icassp40776.2020.9053371</link> ) Array ( [Name] => Copyright [Label] => Rights [Group] => Cpyrght [Data] => CLOSED ) Array ( [Name] => AN [Label] => Accession Number [Group] => ID [Data] => edsair.doi...........858c159b6af41c659715109383e20628 ) |
| RecordInfo |
Array
(
[BibEntity] => Array
(
[Identifiers] => Array
(
[0] => Array
(
[Type] => doi
[Value] => 10.1109/icassp40776.2020.9053371
)
)
[Languages] => Array
(
[0] => Array
(
[Text] => Undetermined
)
)
[Subjects] => Array
(
[0] => Array
(
[SubjectFull] => Similarity (geometry)
[Type] => general
)
[1] => Array
(
[SubjectFull] => Computer science
[Type] => general
)
[2] => Array
(
[SubjectFull] => Speech recognition
[Type] => general
)
[3] => Array
(
[SubjectFull] => Speech synthesis
[Type] => general
)
[4] => Array
(
[SubjectFull] => 010501 environmental sciences
[Type] => general
)
[5] => Array
(
[SubjectFull] => computer.software_genre
[Type] => general
)
[6] => Array
(
[SubjectFull] => 01 natural sciences
[Type] => general
)
[7] => Array
(
[SubjectFull] => Pipeline (software)
[Type] => general
)
[8] => Array
(
[SubjectFull] => 030507 speech-language pathology & audiology
[Type] => general
)
[9] => Array
(
[SubjectFull] => 03 medical and health sciences
[Type] => general
)
[10] => Array
(
[SubjectFull] => End-to-end principle
[Type] => general
)
[11] => Array
(
[SubjectFull] => Transcription (linguistics)
[Type] => general
)
[12] => Array
(
[SubjectFull] => 0305 other medical science
[Type] => general
)
[13] => Array
(
[SubjectFull] => computer
[Type] => general
)
[14] => Array
(
[SubjectFull] => 0105 earth and related environmental sciences
[Type] => general
)
[15] => Array
(
[SubjectFull] => Speaker adaptation
[Type] => general
)
)
[Titles] => Array
(
[0] => Array
(
[TitleFull] => Semi-Supervised Speaker Adaptation for End-to-End Speech Synthesis with Pretrained Models
[Type] => main
)
)
)
[BibRelationships] => Array
(
[HasContributorRelationships] => Array
(
[0] => Array
(
[PersonEntity] => Array
(
[Name] => Array
(
[NameFull] => Sunao Hara
)
)
)
[1] => Array
(
[PersonEntity] => Array
(
[Name] => Array
(
[NameFull] => Katsuki Inoue
)
)
)
[2] => Array
(
[PersonEntity] => Array
(
[Name] => Array
(
[NameFull] => Shinji Watanabe
)
)
)
[3] => Array
(
[PersonEntity] => Array
(
[Name] => Array
(
[NameFull] => Ryuichi Yamamoto
)
)
)
[4] => Array
(
[PersonEntity] => Array
(
[Name] => Array
(
[NameFull] => Masanobu Abe
)
)
)
[5] => Array
(
[PersonEntity] => Array
(
[Name] => Array
(
[NameFull] => Tomoki Hayashi
)
)
)
)
[IsPartOfRelationships] => Array
(
[0] => Array
(
[BibEntity] => Array
(
[Dates] => Array
(
[0] => Array
(
[D] => 01
[M] => 05
[Type] => published
[Y] => 2020
)
)
[Identifiers] => Array
(
[0] => Array
(
[Type] => issn-locals
[Value] => edsair
)
)
[Titles] => Array
(
[0] => Array
(
[TitleFull] => ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)
[Type] => main
)
)
)
)
)
)
)
|
| IllustrationInfo |