SIXTH ANNUAL
ASAP '98 WORKSHOP

Wavelet Based
Optimization of
Space-Time Adaptive
Processing

G. Beylkin
University of Colorado
Boulder, CO

Abstract Currently space-time adaptive array processing is being considered in air-to-air missile applications (e.g. missile intercepts). The challenge for these techniques is the detection and radar-parameter estimation of non-radiating targets in the presence of much higher narrow band, cold clutter and wide band noise generated by low altitude terrain scattered interference (TSI). In representative air-to-air engagements, the number of space-time DOFs required to support a faithful representation of the spatio-temporal-doppler covariance matrix can be large, requiring throughputs that can be prohibitive. For example, simulations of representative scenarios have demonstrated a requirement for approximately 600 DOFs.

The conventional solution to the increased processing requirements of STAP has been faster hardware. The disadvantage of this approach is that dedicated hardware can be costly to develop. Under contract to the DARPA Defense Sciences Office, Raytheon Missile Systems Company, along with the University of Colorado, are investigating a complementary approach of using advanced mathematical techniques, such as wavelets and multi-resolution linear algebras, to optimize and reduce STAP throughput reuqirements. Wavelet filterbanks have been investigated and shown to provide a more compact representation of information in the STAP covariance matrices. In contrast to data dependent transformations such as Karhunen-Loeve, wavelet filterbanks can be precomputed and stored. By sparsening the covariance matrix, STAP computational complexity can be significantly reduced. Wavelet transformations were also constructed that preserve matrix symmetries, such as Block Toeplitz. Wavelets have additional space-time frequency localization properties that may be able to be exploited in STAP.

This talk will focus on theoretical and simulation results of the application of wavelets and non-standard linear algebras to the STAP filtering problem. The current status of program and tools will be presented.