Fast fully adaptive processing: a multistage STAP approach

Due to the need for adequate statistically homogeneous training, full-dimensional space-time adaptive processing (STAP) is well accepted to be impractical. Several previous works have addressed this issue by reducing the adaptive degrees of freedom (DoF), in turn reducing the required training. In this paper, we introduce a new multistage STAP approach that significantly reduces the required sample support while still processing all available DoF. The multistage fast fully adaptive (FFA) scheme draws inspiration from the butterfly structure of the fast Fourier transform. It uses a “divide-and-conquer” approach by creating several smaller STAP problems but then combines the outputs of each problem adaptively as well. The reduction in required sample support rivals currently available reduced DoF algorithms. We also develop three variants of the algorithm, including one that uses random subdivisions of the original STAP problem.We test the efficacy of the algorithms developed via simulations based on simulated airborne radar data and measured high-frequency surface wave radar data. The results show that for simulated homogeneous data, the performance of the FFA approaches is comparable to that of available STAP algorithms; however, with measured data, the FFA approach provides significantly better performance.