Low training volume adaptive processing in HF skywave radar

Adaptive processing in the HF radar environment suffers from the paucity of appropriate training data relative to the dimension of the required spatial filters. We review the applicability of a number of reduced-rank/reduced-order adaptive filtering methods to HF radar and point out the need for several key developments. These developments including properly normalized likelihood ratio for under-sampled scenarios, parametric order estimation based on minimization of both deterministic and stochastic losses, and application of time-varying autoregressive (TVAR) parametric models to the spatially and temporally varying HF radar case.