言語表現による喉頭摘出者のための音声強調システム

• Item type: SIG Technical Reports(1)
• 公開日: 2023-10-07
• タイトル: 言語表現による喉頭摘出者のための音声強調システム
• タイトル / 言語: en
• タイトル: Electrolaryngeal Speech Enhancement Through Strong Linguistic Encoding Methods
• 言語: eng
• キーワード / 主題Scheme: Other
• キーワード / 主題: 音声
• 資源タイプ / 資源タイプ識別子: http://purl.org/coar/resource_type/c_18gh
• 資源タイプ: technical report
• 著者所属: 名古屋大学情報学研究科知能システム学専攻
• 著者所属: 名古屋大学情報学研究科知能システム学専攻
• 著者所属: 名古屋大学情報学研究科知能システム学専攻
• 著者所属: 名古屋大学情報学研究科知能システム学専攻／株式会社TARVO
• 著者所属: 名古屋大学情報基盤センター
• 著者所属(英): en
• 著者所属(英): Graduate School of Informatics, Nagoya University
• 著者所属(英): en
• 著者所属(英): Graduate School of Informatics, Nagoya University
• 著者所属(英): en
• 著者所属(英): Graduate School of Informatics, Nagoya University
• 著者所属(英): en
• 著者所属(英): Graduate School of Informatics, Nagoya University / TARVO, Inc.
• 著者所属(英): en
• 著者所属(英): Information Technology Center, Nagoya University
• 著者名: Lester, Phillip Violeta; Wen-ChinHuang, Ding Ma; 山本, 龍一; 小林, 和弘; 戸田, 智基
• 著者名(英): Lester, Phillip Violeta; Wen-Chin, Huang; Ding, Ma; Ryuichi, Yamamoto; Kazuhiro, Kobayashi; Tomoki, Toda
• 論文抄録 / 内容記述タイプ: Other
• 論文抄録 / 内容記述: Although pretraining and fine-tuning approaches have proven to work well in speech intelligibility enhancement, various mismatches, such as the speech type mismatch or a speaker mismatches between the datasets used in each stage, can deteriorate the conversion performance of this framework. We propose a linguistic encoder robust enough to project both EL and typical speech in the same latent space, while still being able to extract accurate linguistic information, creating a unified representation to reduce the speech type mismatch. Furthermore, we introduce HuBERT output features to the proposed framework for reducing the speaker mismatch. Such a framework makes it possible to effectively use a large-scale parallel dataset during pretraining. We show that compared to the conventional framework using mel-spectrogram input and output features, using the proposed framework enables the model to synthesize more intelligible and naturally sounding speech, as shown by a significant 16% improvement in character error rate and 0.83 improvement in naturalness score.
• 論文抄録(英) / 内容記述タイプ: Other
• 論文抄録(英) / 内容記述: Although pretraining and fine-tuning approaches have proven to work well in speech intelligibility enhancement, various mismatches, such as the speech type mismatch or a speaker mismatches between the datasets used in each stage, can deteriorate the conversion performance of this framework. We propose a linguistic encoder robust enough to project both EL and typical speech in the same latent space, while still being able to extract accurate linguistic information, creating a unified representation to reduce the speech type mismatch. Furthermore, we introduce HuBERT output features to the proposed framework for reducing the speaker mismatch. Such a framework makes it possible to effectively use a large-scale parallel dataset during pretraining. We show that compared to the conventional framework using mel-spectrogram input and output features, using the proposed framework enables the model to synthesize more intelligible and naturally sounding speech, as shown by a significant 16% improvement in character error rate and 0.83 improvement in naturalness score.
• 書誌レコードID / 収録物識別子タイプ: NCID
• 書誌レコードID / 収録物識別子: AN10442647
• 書誌情報: 研究報告音声言語情報処理（SLP）; 巻 2023-SLP-148, 号 8, p. 1-6, 発行日 2023-10-07
• ISSN / 収録物識別子タイプ: ISSN
• ISSN / 収録物識別子: 2188-8663
• Notice: SIG Technical Reports are nonrefereed and hence may later appear in any journals, conferences, symposia, etc.
• 出版者 / 言語: ja
• 出版者: 情報処理学会