A novel hiss noise reduction method for audio signals based on MDCT

A kind of hiss noise reduction method used for audio signal is proposed by Modified Discrete Cosine Transform (MDCT) in this paper. The human auditory model and the parametric soft-thresholding are introduced to the proposed method. A modified median absolute deviation (MAD) is first adopted to avoid overestimate of noise levels. Next, the Modified Discrete Fourier Transform (MDFT) is constructed using the pre-enhanced MDCT coefficients and the masking threshold and the adaptive masking parameters are calculated in MDFT domain. Finally, a parametric soft-thresholding method is employed to attenuate hiss noise significantly and keep more high-frequency (HF) information. The objective and subjective listening test results show that the proposed algorithm outperforms the referenced methods.