Cramer-Rao Bounds for Underdetermined Source Localization

Although Cramer-Rao Bounds (CRB) for direction-of-arrival (DOA) estimation have been extensively studied for decades, existing results are mainly applicable when there are fewer sources than sensors. In contrast, this letter considers an underdetermined signal model (more sources than sensors) and investigates conditions under which CRB exist. Necessary and sufficient conditions are derived for the associated Fisher information matrix to be nonsingular, which in turn, leads to closed-form expressions for the CRBs for underdetermined DOA estimation. These conditions highlight crucial roles played by the array geometry, as well as the correlation between source signals. The CRB for different array geometries are numerically compared both in the overdetermined and underdetermined settings.