Data reduction in sensor array processing using parameterized signals observed in colored noise

This paper addresses the issue of adaptive optimal beamspace transformation in sensor array processing. Assuming a deterministic, parameterized signal model, and spatially colored, but temporally white Gaussian noise and interference, an optimality condition on how to choose the beamspace transformation is derived. This condition ensures that the Cramer-Rao bounds of the parameter estimates are preserved by the transformation. The optimal transformation depends on the unknown array correlation matrix and the directions of arrival. In view of this fact, practical procedures for approximating the optimal transformation are discussed. Simulations that support the theoretical results are included for the case where the source signals are sinusoidal.