Using noise statistics for effective noise filtering

In this paper we show that the knowledge of noise statistics contaminating a signal leads to a better choice of filter to remove the noise. Very specifically, we show theoretically that the additive white Gaussian noise (AWGN) contaminating a signal can be filtered best by using a Gaussian filter mask which has some relation with the noise statistic of the AWGN. The main contribution of the paper is (a) the derivation of the relationship between the Gaussian mask and the noise statistics and (b) demonstration of its effective use in speech recognition.