In this paper, we propose a family of Bayesian estimators for single channel speech enhancement. Three different kinds of enhancement estimators are derived from the statistics of bivariate Gamma model. In the Gamma model, it is assumed that the spectral components are correlated with each other. The experimental results show that the proposed estimators are effective for speech enhancement compared to state-of-the-art estimators.

Index Terms: speech enhancement, bivariate Gamma model, Gaussian distribution, correlated spectral components