Training method in space-time adaptive processing with strong generalizing ability

The effects of array errors on the clutter spectra are discussed. It is pointed out that, just like that in pattern recognition problems, the training method has significant effects on the performance of space-time adaptive processing (STAP) techniques. A reasonable training method is given, which exhibits strong generalizing ability. Simulation results based on both the steady-state clutter covariance matrix and the high fidelity radar clutter data confirm the above conclusions.