Adaptive space-time processing for radar receive arrays with time-varying two-dimensional subband decompositions

A new two-dimensional adaptive processor structure is introduced for radar receive arrays. This new processor utilizes a tree-structured filterbank to implement a non-uniform partitioning of the spatio-temporal frequency domain, and thus provides a multiresolution approach to jammer localization and adaptive nulling. It is shown that when platform restrictions limit the allowable number of adaptive weights, the filterbank structure processor utilizing the simple normalized LMS algorithm is capable of a faster speed of convergence than the standard tapped-delay-line structure employing the LMS or least squares algorithms. Furthermore, it is possible for the minimum mean-square error of the filterbank processor to be lower than that attainable with the standard tapped-delay-line structure.<>