Paper Title
Student -T prior based Single Channel Speech Enhancement using Iterative Posterior NMF

Abstract
This paper proposes a speech enhancement technique based on regularized NMF (non-negative matrix factorization) for non-stationary Gaussian noise. In an iterative posterior NMF-based model, the magnitudes of spectral elements of the noise and speech are considered as prior distributions. This is used since the sample distributions of the above are well suited to Student T and Rayleigh densities well. For the time-varying noise environments, both the speech and noise bases of NMF are adapted simultaneously. This paper proposes to adaptively estimate the statistics of these distributions. The proposed method gives the better results in terms of perceptual evaluation of speech quality (PESQ) and the signal-to-distortion ratio (SDR). Keywords - Student T Probability Density Function, PSEQ, Signal-to-Distortion Ratio, Posterior Regularization, Non-Negative Matrix Factorization (NMF)