|
JAIST Repository >
School of Information Science >
JAIST Research Reports >
Research Report - School of Information Science ： ISSN 0918-7553 >
IS-RR-2005 >
Please use this identifier to cite or link to this item:
http://hdl.handle.net/10119/8404
|
| Title: | Fundamental frequency estimation for noisy speech based on instantaneous amplitude and frequency |
| Authors: | Ishimoto, Yuichi
Unoki, Masashi
Akagi, Masato |
| Issue Date: | 2005-03-28 |
| Publisher: | 北陸先端科学技術大学院大学情報科学研究科 |
| Magazine name: | Research report (School of Information Science, Japan Advanced Institute of Science and Technology) |
| Volume: | IS-RR-2005-006 |
| Start page: | 1 |
| End page: | 31 |
| Abstract: | This paper proposes a robust and accurate method of estimating the fundamental frequencies (F0s) for noisy speech. In general, it is difficult to directly estimate accurate F0s from noisy speech. This method combines two different methods of F0 estimation. One is based on the periodicity and harmonicity of instantaneous amplitude of speech; it is robust against noise, but it does not allow for accurate F0 estimation. The other is based on the stability of instantaneous frequency, and it enables accurate F0 estimation, but this method is not robust against noise. To combine these two methods, the proposed method makes use of noise reduction by using a comb filter with controllable pass-bands. Experiments were carried out to estimate F0s of real speech in noisy environments and to compare the proposed method with other methods such as an autocorrelation methods and a cepstrum method. The results showed that this method was more robust than the other methods. This method could estimate F0s of noisy speech with accuracy similar to that in clean speech F0 estimation by using only the stability of instaneous frequency. |
| URI: | http://hdl.handle.net/10119/8404 |
| Material Type: | publisher |
| Appears in Collections: | IS-RR-2005
|
Files in This Item:
| File |
Description |
Size | Format |
|---|
| IS-RR-2005-006.pdf | | 710Kb | Adobe PDF | View/Open |
|
All items in DSpace are protected by copyright, with all rights reserved.
|
