Space-Time Adaptive Processing Using Pattern Classification

Since it was first developed to solve the problem of target detection by moving target indicator (MTI) radars, space-time adaptive processing (STAP) has seen many versions, developed to overcome the shortcomings of the original version. In this paper, we introduce a new method, called Learning-Based Space-Time Adaptive Processing (LBSTAP), in which the detection problem is approached from the point of view of classification. It is shown that the proposed technique offers an advantage over STAP in terms of output SINR in cases where the amount of training data is limited and the signal-to-interference ratio is higher than -20 dB. Moreover, it is shown that LBSTAP is more resilient to clutter variations and the problem of target cancellation. A cascaded system of STAP followed by LBSTAP is also introduced to enhance the performance of LBSTAP in cases of low-power targets. The cascaded system is shown to outperform both individual systems, albeit at the price of higher computational complexity.