Overview

The optimum space-time processor as analysed in Chapter 4 suffers from a high degree of computational complexity so that its use in practical applications, especially for real-time operation, is unlikely.

Transforming the covariance matrix by the discrete wavelet transform (DWT) may be a way of reducing the computational load for matrix inversion, exploiting the sparsening property of the DWT (BRAUNREITER et al. [52]).

In Chapters 5, 6, 7 and 9 we discuss several ways of reducing the signal vector space and, hence, the computational workload associated with space-time adaptive MTI processing. All of these techniques are based on linear subspace transforms as has been described already in Section 1.2.3.

The principle of linear transforms to reduce the signal subspace ^[1] has been addressed by several authors (KLEMM [240], WARD [530, p. 81], HAIMOVICH et al. [194], SELIKTAR et al. [459], BARANOSKI [28]). WANG Y. and PENG [524] present a unified approach for various kinds of transform processors (element-pulse, beamspace-pulse, element-Doppler, beamspace-Doppler). A similar overview of these techniques has been given by WARD [531, 530].

BOJANCZYK and MELVIN [46] analyse a least squares STAP technique based on the preconditioned conjugate gradients iterative method.

HIMED and MELVIN [208] give a brief overview of different subspace processors: factored times-space (frequency-dependent spatial processing), extended factored time-space (frequency dependent spatial processing, with additional auxiliary Doppler channels involved), adaptive displaced phase centre (ADCPA), eigencanceller (orthogonal projection), and eigen-based cross-spectral metric (CSM, GOLDSTEIN and REED [160], GUERCI et al. [187]). Some...