WEKO3
アイテム
音韻知識にバイアスされた音韻推定：PHONOBEST～現状と残された課題～
https://doi.org/10.15100/0000014986
https://doi.org/10.15100/00000149864836bef5-12f0-46a1-b0cb-05cb598c8479
| 名前 / ファイル | ライセンス | アクション |
|---|---|---|
AA12013209-20071130-0009.pdf (1.3 MB)
|
|
| Item type | ☆紀要論文 / Departmental Bulletin Paper(1) | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 公開日 | 2016-06-01 | |||||||||||
| タイトル | ||||||||||||
| タイトル | 音韻知識にバイアスされた音韻推定：PHONOBEST～現状と残された課題～ | |||||||||||
| 言語 | ja | |||||||||||
| タイトル | ||||||||||||
| タイトル | PHONOBEST: Phonetic Knowledge biased Estimation～Current Status and Future Works～ | |||||||||||
| 言語 | en | |||||||||||
| 著者 |
勝瀬, 郁代
× 勝瀬, 郁代
|
|||||||||||
| 言語 | ||||||||||||
| 言語 | jpn | |||||||||||
| キーワード | ||||||||||||
| 主題 | 自動音声認識, 音韻知識, トップダウン処理, 音声知覚モデル, 調波構造 automatic speech recognition, phonetic knowledge, top-down process, speech perception model, harmonic structure |
|||||||||||
| 資源タイプ | ||||||||||||
| 資源タイプ識別子 | http://purl.org/coar/resource_type/c_6501 | |||||||||||
| 資源タイプ | departmental bulletin paper | |||||||||||
| ID登録 | ||||||||||||
| ID登録 | 10.15100/0000014986 | |||||||||||
| ID登録タイプ | JaLC | |||||||||||
| 版 | ||||||||||||
| 出版タイプ | NA | |||||||||||
| 出版タイプResource | http://purl.org/coar/version/c_be7fb7dd8ff6fe43 | |||||||||||
| 出版者 名前 | ||||||||||||
| 出版者 | 近畿大学産業理工学部 | |||||||||||
| 書誌情報 |
ja : かやのもり：近畿大学産業理工学部研究報告 en : Reports of Faculty of Humanity-Oriented Science and Engineering, Kinki University 巻 7, 号 2, p. 9-17, 発行日 2007-11-30 |
|||||||||||
| ISSN | ||||||||||||
| 収録物識別子タイプ | ISSN | |||||||||||
| 収録物識別子 | 13495801 | |||||||||||
| 内容記述 | ||||||||||||
| 内容記述タイプ | Abstract | |||||||||||
| 内容記述 | A new automatic speech recognition (ASR) method, PHONOBEST, is proposed, aiming to achieve robust ASR in a real environment. In ASRs, we need to compare observed speech with template speech in the pattern matching process. Such template speech can be regarded as a kind of top-down information generated from speech schema. In human speech perception, the top-down process seems to work more than preparing candidates for pattern matching. It also seems to predispose the segregation process to extract an object agreeing with the "expectation" generated by speech schema. PHONOBEST has introduced this active bias into the process of target speech estimation. A missingdata model is also implemented in PHONOBEST by using fine spectral representation with harmonic structures as template patterns; on the other hand, the spectral envelopes are supplied as template patterns in the prevailing ASRs. In this paper, the current status and remaining assignments are discussed. | |||||||||||
| 言語 | en | |||||||||||
