Publication detail

Power Spectral Density Noise Estimation using Wavelet Transform in Spectral Domain

SYSEL, P. SMÉKAL, Z.

Original Title

Power Spectral Density Noise Estimation using Wavelet Transform in Spectral Domain

Type

conference paper

Language

English

Original Abstract

The paper describes the design of a new method of power spectral density estimation using the wavelet transform in the spectral domain. To reduce periodogram variance the proposed method uses the procedure of thresholding the wavelet coefficients of a periodogram. The periodogram is decomposed into a chosen number of levels using the discrete wavelet transform with the appropriate mother wavelet. Then the thresholds used for the wavelet transform coefficients in dependence on the level are calculated. To threshold the wavelet transform coefficients soft thesholding is used. Then the smoothed estimate of power spectral density of noise is obtained using the inverse discrete wavelet transform. This reduces the variance of the estimate of power spectral density of noise.

Keywords

power spectral density, estimation bias, estimation variance, spectral subtraction

Authors

SYSEL, P.; SMÉKAL, Z.

RIV year

2007

Released

21. 9. 2007

Location

Praha

ISBN

978-80-86269-00-9

Book

Proceedings of 17th Czech-German Workshop Speech Processing

Pages from

98

Pages to

105

Pages count

8

BibTex

@inproceedings{BUT25587,
  author="Petr {Sysel} and Zdeněk {Smékal}",
  title="Power Spectral Density Noise Estimation using Wavelet Transform in Spectral Domain",
  booktitle="Proceedings of 17th Czech-German Workshop Speech Processing",
  year="2007",
  pages="98--105",
  address="Praha",
  isbn="978-80-86269-00-9"
}