Optimal partially adaptive sensor array processing

The problem addressed in this paper is minimum variance adaptive sensor array processing subject to limitations on the dimension of the adaptive processor. Advances in technology have made it possible for space-segment and airborne platforms to support arrays composed of many elements for communications and radar systems. However, the computational complexity requirements of such sensor arrays, coupled with the desire or requirement for space-time processing, may prohibit full adaptivity. A new technique for rank reduction based upon a cross-spectral performance index is introduced, and it is shown that this method results in a lower minimum mean-square error (MMSE) than the principal components method of rank reduction. An example is provided which demonstrates that the Wiener filter operating in the subspace selected by this new metric outperforms the optimal filter operating in the subspace chosen based upon the largest eigenvalues criteria.