|
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/10119/4891
|
| タイトル: | Estimation of local peaks based on particle filter in adverse environments |
| 著者: | Tomoike, Seiji
Akagi, Masato |
| 発行日: | 2008-07 |
| 出版者: | Research Institute of Signal Processing Japan（信号処理学会） |
| 誌名: | Journal of Signal Processing |
| 巻: | 12 |
| 号: | 4 |
| 開始ページ: | 303 |
| 終了ページ: | 306 |
| 抄録: | In this paper, we propose an estimation method for local peaks of the speech spectrum using a particle filter in noisy environments. The conventional local peak estimation methods do not use the estimated peak knowledge in previous frames. These methods estimate the local peaks only in the current frame. Therefore, noises markedly affect the accuracy of local peak estimation in the current frame. The sudden incidence of peak candidates that are not appropriate for local peaks might be estimated in noisy environments. Thus, these methods have to robustness against nonstationary noise. To solve the problems of conventional peak-picking methods, we proposed a two-step estimation method for local peaks using a particle filter. The first step is to estimate the peak presence probability on the basis the spectral envelope of the cepstrum. The local peaks can be simultaneously estimated using likelihood with the same peak presence probability in the high-probability regions. The second step is to extract peaks from the candidates of the peaks on the basis of the peak presence probability. Experimental results show that the proposed method is superior to conventional methods in terms of the frequency distance and the number of correct peaks in the nonstationary noisy environment. |
| Rights: | Copyright (C) 2008 Research Institute of Signal Processing Japan. Seiji Tomoike and Masato Akagi, Journal of Signal Processing, 12(4), 2008, 303-306. |
| URI: | http://hdl.handle.net/10119/4891 |
| 資料タイプ: | publisher |
| 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 |
|---|
| JSP_Tomoike_2008.pdf | | 313Kb | 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。
|
