Dual-channel optimally modified log-spectral amplitude estimator using spatial information

In this paper, we present a dual channel speech enhancement algorithm using optimally modified log-spectral amplitude estimator based on spatial information, which minimizes the mean-square error of the log-spectral for speech signals under target presence uncertainty. A target presence probability estimator based on the phase difference and magnitude squared coherence information to distinguish signals coming from the desired and undesired directions in each frequency bin is proposed. Experimental results show that the proposed algorithm achieves significantly reduced level of nonstationary noise without further distorting the desired signal components over the comparative algorithms.