Title :
A Speech Enhancement Algorithm Based on Bark-Scale Wavelet Package
Author :
Teng, Jian-Fu ; Dong, Jian ; Wang, Shu-Yan ; Bao, Hu ; Wang, Ming-Guo
Author_Institution :
Tianjin Univ. of Commerce, Tianjin
Abstract :
The selection of threshold and curse is a key issue to determine the de-noisy effect and distortion of the output signal in speech enhancement algorithm. In this paper, combined with Bark-scale Wavelet Package, the S curve of QNN (Quantum Neural Networks) is introduced to wavelet threshold method to achieve speech enhancement. The simulation result indicates that the method presented is superior to that of traditional soft or hard threshold method. The algorithm presented can improve the quality of voice.
Keywords :
distortion; neural nets; signal denoising; speech enhancement; wavelet transforms; bark-scale wavelet package; quantum neural networks; signal denoising; signal distortion; speech enhancement; wavelet threshold; Frequency; Machine learning; Machine learning algorithms; Packaging machines; Signal processing; Signal processing algorithms; Speech enhancement; Wavelet analysis; Wavelet domain; Wavelet transforms; QNN; Speech enhancement; Wavelet package;
Conference_Titel :
Machine Learning and Cybernetics, 2007 International Conference on
Conference_Location :
Hong Kong
Print_ISBN :
978-1-4244-0973-0
Electronic_ISBN :
978-1-4244-0973-0
DOI :
10.1109/ICMLC.2007.4370733
