Optimal Filters in the KLE Domain

We have seen the advantages of transforming the noisy speech into the frequency domain and then performing noise reduction. Although it is the most studied one, the frequency domain is not the only domain in which we can design interesting filters for noise reduction. There are other domains that may offer more advantages in the filter design. In this chapter, we consider the KLE domain. we derive two broad classes of optimal filters in this domain depending basically on the subband filter length. The first class, similar to the frequency-domain filters, estimates a frame of the clean speech by filtering the corresponding frame of the noisy speech while the second class does noise reduction by filtering not only the current frame but also a number of previous consecutive frames of the noisy speech.