Speech Enhancement in the Dft Domain Using Laplacian Speech Priors

In this paper we consider optimal estimators for speech enhancement in the Discrete Fourier Transform (DFT) domain. We derive an analytical solution for estimating complex DFT coefficients in the MMSE sense when the clean speech DFT coefficients are Laplacian distributed and the DFT coefficients of the noise are Gaussian or Laplacian distributed. We show that these estimators have a number of interesting properties. Compared to previously proposed estimators, which are based on Gamma speech priors, the estimators based on Laplacian speech priors have a simpler analytic form.