期刊名称：International Journal on Smart Sensing and Intelligent Systems
印刷版ISSN：1178-5608
出版年度：2015
卷号：8
期号：4
页码：2231-2246
出版社：Massey University
摘要：S-transform (ST) is an effective time-frequency representation method with the advantages ofshort-time Fourier transform and wavelet transform. This paper utilizes the advantages of the powerspectral subtraction method and the speech enhancement method to presents a novel speechenhancement algorithm based on discrete ST. Firstly, the speech in time domain is transformed by STto joint time-frequency domain for spectral subtraction so that the clear speech spectrum can beobtained and then the inverse ST is performed to acquire the enhanced speech in time domain.Simulation experiment is done to verify the validity of the method. The experiment results show that theproposed method can effectively enhance the de-noising ability and improve the signal-to-noise ratio.
