|
JAIST Repository >
b. 情報科学研究科・情報科学系 >
b10. 学術雑誌論文等 >
b10-1. 雑誌掲載論文 >
このアイテムの引用には次の識別子を使用してください:
http://hdl.handle.net/10119/4900
|
| タイトル: | A noise reduction system based on hybrid noise estimation technique and post-filtering in arbitrary noise environments |
| 著者: | Li, Junfeng
Akagi, Masato |
| キーワード: | Hybrid Noise Estimation
Post-Filtering
Coherence Function
Speech Presence Uncertainty |
| 発行日: | 2006-02 |
| 出版者: | Elsevier |
| 誌名: | Speech Communication |
| 巻: | 48 |
| 号: | 2 |
| 開始ページ: | 111 |
| 終了ページ: | 126 |
| DOI: | 10.1016/j.specom.2005.06.013 |
| 抄録: | In this paper, we propose a novel noise reduction system using a hybrid noise estimation technique and post-filtering which is effective in dealing with localized and non-localized noise components. To suppress localized noises, we develop a hybrid noise estimation technique that combines a reformulated multi-channel estimation technique and a soft-decision single-channel estimation technique. Final estimation accuracy is significantly improved by a speech absence probability estimator which considers the strong correlations of speech presence uncertainty between adjacent frequency components and consecutive frames. The estimated spectra of localized noises are subtracted from those of noisy observations by spectral subtraction. The non-localized noises are further reduced by a multi-channel post-filter which minimizes the mean squared error of log-spectra. An estimator for the a priori speech absence probability is derived from the coherence characteristic of the noise field at spectral subtraction output, improving the spectral enhancement of the desired speech signal. Experimental results demonstrated that the hybrid estimation technique produces more accurate spectral estimates for localized noises and the proposed noise reduction system results in significant improvements in terms of objective speech quality measures: segmental SNR (SEGSNR) and Mel-Frequency Cepstral Coefficient (MFCC) distance and subjective speech quality measures: spectrograms and listening tests in various noise conditions. |
| Rights: | NOTICE: This is the author's version of a work accepted for publication by Elsevier. Junfeng Li and Masato Akagi, Speech Communication, 48(2), 2006, 111-126, http://dx.doi.org/10.1016/j.specom.2005.06.013 |
| URI: | http://hdl.handle.net/10119/4900 |
| 資料タイプ: | author |
| 出現コレクション: | b10-1. 雑誌掲載論文 (Journal Articles)
|
このアイテムのファイル:
| ファイル |
記述 |
サイズ | 形式 |
|---|
| C6969.pdf | | 487Kb | Adobe PDF | 見る/開く |
|
当システムに保管されているアイテムはすべて著作権により保護されています。
|
