Real-time whitening application to two microphone sensors for comb filtering and smoothing

This paper presents a real-time comb filtering and smoothing technique on two-microphone sensors by applying a whitening filter and a prediction filter on the sum and subtract functions. The analysis shows that from the sum and subtract functions, each output signals from a cascaded primary input after the sum function and a cascaded reference input after the subtract function give a favorable effect of a comb filtering and a linear prediction respectively. Furthermore, based on the fact that a minimum phase whitening filter produces a whitened signal with a flat spectrum, a minimum phase whitening filter and a prediction filter are applied before a subtract function and also after a sum function, respectively. Therefore, it will be shown in beamforming structure that the whitened based signal gives a favorable effect of a comb filtering, and also a smoothing on an application of an adaptive filter together.