In this paper, utilization of wavelet filters via multistage convolution has been investigated using Reverse Biorthogonal Wavelets (RBW). The speech signal portions of frequencies are decomposed in high and low pass bands filters and then the noise is removed separately by the implementation of wavelet filters at different stages. Using this approach we avoid the use of conventional thresholding that affect the information contained in the speech signal. This method gives better results and outcomes. In order to evaluate the presented method, different objective measures using different noise probability distribution functions are implemented and it is also tested by subjective evaluating. Comparing RBW method with Donoho and Johnstone thresholding method and Birge-Massart thresholding strategy method, it showed better and more effect results.
Track
CCAT: Communications and Computer Applications and Technologies
Conference
3rd Mosharaka International Conference on Communications, Computers and Applications (MIC-CCA 2009)
Congress
2009 Global Congress on Communications, Computers and Applications (GC-CCA 2009), 26-28 October 2009, Amman, Jordan
Pages
15-27
Topics
Wavelet Processing Digital Signal Processing Digital Speech Processing
ISSN
2227-331X
DOI
BibTeX
@inproceedings{47CCA2009,
title={The study of wavelet filter speech enhancement method},
author={Khaled Daqrouq, and Abdel-Rahman K. Al-Qawasmi},
booktitle={2009 Global Congress on Communications, Computers and Applications (GC-CCA 2009)},
year={2009},
pages={15-27},
doi={}},
organization={Mosharaka for Research and Studies}
}