Improved priori SNR estimation for sound enhancement with Gaussian statistical model

In allusion to the estimation problem of priori signal to noise ratio parameter in sound enhancement, by using Gaussian statistical model, a novel estimation algorithm for a priori signal to noise ratio was proposed in frequency domain. The presented algorithm with MMSE (Minimum Mean Square Error) computes directly the spectrum of the clean sound component to obtain the estimated priori signal to noise ratio of the decision directed approach, and thus the shortcoming of the two step noise reduction method is effectively eliminated. Moreover, this algorithm has good excellent performance in highly reducing the noise of the output sound while the advantages in noise suppression are retained. Comparing with the classic decision directed method and the recently proposed two step noise reduction technique, experimental results show that the proposed method in this paper has excellent performance under different noise background.