Broadband signal subspace processing for range estimation

The problem of broadband range and bearing estimation is considered. Coherent focusing of the data covariance matrix over multiple frequency bins for 2D MUSIC estimation is studied and compared to incoherent summing of the MUSIC cost function over frequency bins. Coherent focusing provides improved performance over incoherent processing in the farfield bearing-only estimation problem, but introduces instabilities when applied to the near-field range and bearing estimation problem. The cost function response in the range parameter is not highly peaked, so initial focusing using even a slightly incorrect range estimate prevents convergence of the range estimate to the true range value. Incoherent processing, on the other hand, is less computationally complex, and achieves the Cramer-Rao lower bound at high SNR.