Linear predeictive resolution of uncorrelated and correlated signals

We investigate the resolution and bias of two linear prediction (LP) algorithms for bearing estimation with linear array of sensors, in the presence of incoherent and coherent sources: the Burg algorithm (BA) and the generalized Burg algorithm (GBA). The superresolution is easily achieved with both algorithms when the sources are incoherent. For coherent sources, however, the BA estimate is somewhat more sensitive to the magnitude and the phase of the inter-signal correlation than the GBA, thereby resulting in the increased bias of the bearing estimates.