Maximum likelihood and robust G-music performance in K-distributed noise

Abstract

For an antenna array input mixture of m point source signals in K-distributed noise, we compare DOA estimation delivered by Maximum Likelihood and the recently introduced Robust G-MUSIC (RG-MUSIC) technique. We demonstrate that similar to the Gaussian case, MLE is still superior to RG-MUSIC, especially within the so-called threshold region. This makes it possible to use the Expected Likelihood (EL) methodology to detect the presence of RG-MUSIC performance breakdown and "cure" those cases via an approach previously developed for the complex Gaussian circumstance.