講演抄録/キーワード |
講演名 |
2013-07-19 15:45
Semi-Blind Optimization Scheme of Joint Suppression of Background Noise and Late Reverberation ○Fine D. Aprilyanti・Hiroshi Saruwatari・Kiyohiro Shikano・Satoshi Nakamura(NAIST)・Tomoya Takatani(Toyota) EA2013-52 |
抄録 |
(和) |
(まだ登録されていません) |
(英) |
Recently, a method to jointly suppress diffuse background noise and late reverberation part of speech has been proposed, combining frequency domain speech signal extraction and Wiener filter. In this paper, we propose an automatic optimization scheme of this joint method to improve the speech recognition accuracy for spoken-dialogue robot. We optimize the internal parameter of the conventional method using the assessment of musical noise measured by higher-order statistics and acoustic model likelihood. In addition, we apply the switching scheme using the distance information provided by image sensor to maintain the optimum performance of this method. The experimental results show that the proposed approach improves the word recognition accuracy. |
キーワード |
(和) |
/ / / / / / / |
(英) |
Blind signal extraction / noise suppression / dereverberation / higher-order statistics / acoustic likelihood / / / |
文献情報 |
信学技報, vol. 113, no. 134, EA2013-52, pp. 111-116, 2013年7月. |
資料番号 |
EA2013-52 |
発行日 |
2013-07-11 (EA) |
ISSN |
Print edition: ISSN 0913-5685 Online edition: ISSN 2432-6380 |
著作権に ついて |
技術研究報告に掲載された論文の著作権は電子情報通信学会に帰属します.(許諾番号:10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
PDFダウンロード |
EA2013-52 |