Subband STAP Processing, the Fifth Generation
Allan Steinhardt
Adaptive array processing has moved through several stages of evolution. First there was spatial-only adaptation, using LMS gradient-based algorithms. Second was true space-time adaptivity using time taps on multiple channels.Third came more rapid convergence, using sample matrix inversion strategies. Fourth came dimensionally large system solutions, which focused on the need for reduced DOF such as beamspace, eigenspace, MUSIC, ESPRIT, BASALE, etc. This paper argues that we are on the threshold of the fifth generation of STAP; mainly STAP in the subband domain. Subband STAP, or SSTAP, is an elegant and computationally efficient solution to the need for increasing bandwidth in radar and sonar processing as well as mobile communications. We will begin by motivating the need for wider bandwidths, from an ATR sensing perspective as well as multimedia communications. Next we will show that even with advanced eigenmethods the computational burden of wideband STAP is untenable for existing and emerging military requirements.
Having motivated a compelling need, we then propose SSTAP as a solution, motivated from an active sensor perspective. We will describe SSTAP as an extension of the familiar STAP 3-D data cube (range, angle, Doppler) to a 4-D Hypercube (subband range, angle, subband Doppler, subband). Subbanding allows us to overcome challenging wideband effects such as Doppler-frequency coupling, high resolution range migration, target self nulling, channel decorrelation, and dispersion. We will describe the challenges of SSTAP interference rejection and subband recombining with artifacts, and present a processing architecture for a surveillance radar example.