Analytical Performance Assessment of 1-D ESPRIT in DFT Beamspace in Terms of Physical Parameters.

In this paper, we present an asymptotic performance analysis of a subspace-based parameter estimation scheme in DFT beamspace that is based on a first-order expansion of the estimation error that is due to additive noise. We provide a general expression of the mean squared error (MSE) for 1-D ESPRIT in DFT beamspace as a function of the perturbation in terms of physical parameters, such as the array steering matrix, the beamforming matrix, the signal correlation, the number of snapshots, and the size of the array. Additionally, a closed-form expression is provided for the case of a single source. This paper includes the main parts of the derivation. For the simulation results, we compare the analytical performance with the empirical one that is based on Monte-Carlo trials. We also compare the performance of the DFT beamspace version with the element space counterpart of the ESPRIT algorithm and reveal the benefits of the former. In conclusion, we discuss possible extensions for follow-up investigations.