I-vector Domain Adaptation Using Cycle-Consistent Adversarial Networks for Speaker Recognition

• Item type: SIG Technical Reports(1)
• 公開日: 2019-02-20
• タイトル: I-vector Domain Adaptation Using Cycle-Consistent Adversarial Networks for Speaker Recognition
• タイトル / 言語: en
• タイトル: I-vector Domain Adaptation Using Cycle-Consistent Adversarial Networks for Speaker Recognition
• 言語: eng
• 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: Tokyo Institute of Technology
• 著者所属: Tokyo Institute of Technology
• 著者所属(英): en
• 著者所属(英): Tokyo Institute of Technology
• 著者所属(英): en
• 著者所属(英): Tokyo Institute of Technology
• 著者名: Yi, Liu; Takahiro, Shinozaki
• 著者名(英): Yi, Liu; Takahiro, Shinozaki
• 論文抄録 / 内容記述タイプ: Other
• 論文抄録 / 内容記述: Speaker recognition systems often suffer from severe performance degradation due to the difference between training data and evaluation data, which is called domain mismatch problem. In this paper, we apply adversarial strategies in deep learning techniques and propose a method using cycle-consistent adversarial networks for i-vector domain adaptation. This method performs an i-vector domain transformation from the source domain to the target domain to reduce the domain mismatch. It uses a cycle structure that reduces the negative influence of losing speaker information in i-vector during the transformation and makes it possible to use unpaired datasets for training. The experimental results show that the proposed adaptation method improves recognition performance of a conventional i-vector and PLDA based speaker recognition system by reducing the domain mismatch between the training and the evaluation sets.
• 論文抄録(英) / 内容記述タイプ: Other
• 論文抄録(英) / 内容記述: Speaker recognition systems often suffer from severe performance degradation due to the difference between training data and evaluation data, which is called domain mismatch problem. In this paper, we apply adversarial strategies in deep learning techniques and propose a method using cycle-consistent adversarial networks for i-vector domain adaptation. This method performs an i-vector domain transformation from the source domain to the target domain to reduce the domain mismatch. It uses a cycle structure that reduces the negative influence of losing speaker information in i-vector during the transformation and makes it possible to use unpaired datasets for training. The experimental results show that the proposed adaptation method improves recognition performance of a conventional i-vector and PLDA based speaker recognition system by reducing the domain mismatch between the training and the evaluation sets.
• 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 2019-SLP-126, 号 2, p. 1-3, 発行日 2019-02-20
• 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