Statistical analysis of a subspace method for bearing estimation without eigendecomposition

The paper studies the statistical properties of a subspace-based method for bearing estimation without eigendecomposition. For easy reference, we use the acronym BEWE (bearing estimation without eigendecomposition) for the method. The BEWE large-sample variance is derived and shown to be bounded from below by the MUSIC large-sample variance. The drawback of being less accurate than MUSIC is balanced by BEWE's computational advantage. In addition, it is shown that, unlike MUSIC, BEWE can accommodate the case of spatially finitely-correlated sensor noise.