Research for unsupervised acoustic pattern discovery with zero resources
| Project/Area Number |
17K00237
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Multi-year Fund |
|---|
| Section | 一般 |
|---|
| Research Field |
Perceptual information processing
|
|---|
| Research Institution | Nara Institute of Science and Technology |
|---|
Principal Investigator |
Sakti Sakriani 奈良先端科学技術大学院大学, 先端科学技術研究科, 特任准教授 (00395005)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
中村 哲 奈良先端科学技術大学院大学, データ駆動型サイエンス創造センター, 教授 (30263429)
|
|---|
| Project Period (FY) |
2017-04-01 – 2020-03-31
|
| Project Status |
Completed (Fiscal Year 2019)
|
| Budget Amount *help |
¥4,550,000 (Direct Cost: ¥3,500,000、Indirect Cost: ¥1,050,000)
Fiscal Year 2019: ¥1,040,000 (Direct Cost: ¥800,000、Indirect Cost: ¥240,000)
Fiscal Year 2018: ¥1,560,000 (Direct Cost: ¥1,200,000、Indirect Cost: ¥360,000)
Fiscal Year 2017: ¥1,950,000 (Direct Cost: ¥1,500,000、Indirect Cost: ¥450,000)
|
|---|
| Keywords | 音声認識 / ゼロ資源音声技術 / 脳波 / 音声翻訳 / 音声情報処理 |
|---|
| Outline of Final Research Achievements |
With the Tokyo Olympics and Paralympics approaching, language barriers between tourists are becoming critical problems to overcome. Current speech recognition and speech translation have been readily available, but only for several languages where large resources are available. Here, we addressed zero-resource speech problem where language specific knowledge and collection of transcribed data are not available. In order to understand the unknown language, we analyzed and investigated the process by which the human brain processes language. In addition, we have developed a closed-loop speech chain model based on deep learning so that we can learn how to listen while the machine is speaking. This is the first deep learning model that integrates human speech recognition and production behavior.
|
| Academic Significance and Societal Importance of the Research Achievements |
アフリカ言語（ツォンガ語）とインドネシア言語のゼロリソースモデリングの構築に成功した。また、2017年と2019年の世界ゼロ資源スピーチチャレンジに参加し、提案手法で上位結果を得ることができた。さらに、深層学習に基づく閉ループスピーチチェインモデルを開発して、機械が話している間、聞く方法を学習できるようにした。2019年では世界言語言語コンソーシアムのためにユネスコとも協力した。この研究の結果は、トップ会議（ASRU、Interspeech、ICASSP）とトップジャーナル(IEEE / ACM TASLP)で公開された。さらに、スピーチチェインモデルの特許も取得した。
|
Report
(4 results)
Research Products
(77 results)
-
-
-
-
-
[Journal Article] Machine Speech Chain2020
Author(s)
Andros Tjandra, Sakriani Sakti, Satoshi Nakamura
-
Journal Title
IEEE/ACM Transactions on Audio, Speech, and Language Processing
Volume: －
Pages: 976-989
DOI
Related Report
Peer Reviewed / Open Access / Int'l Joint Research
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
