A new filterbank-based MMSE approach to reducing the degrees of freedom in adaptive sensor array processing

This paper introduces a cross-spectral minimum-mean square error approach for the reduction in degrees of freedom in partially adaptive array processing. In space communications, as with all space-segment and airborne sensors, size-weight-power platform constraints coupled with the desire or requirement for large element arrays and space-time adaptive processing make fully adaptive array processors unrealizable. The general solution to reducing the degrees of freedom for array processors is addressed from a joint-process point of view using the geometry of least squares to select a lower dimensional subspace for adaptive nulling. It is demonstrated that the subspace selection based on this technique may outperform the principal components method. A tree-structured time-varying filterbank is proposed to provide an adaptive transformation for the subspace selection