Power selected Training
for False Alarm Mitigation in
|Daniel J. Rabideau and Allan O. Steinhardt
MIT Lincoln Laboratory
244 Wood Street
Lexington, MA 02173-9108
tel: (617) 981-2892
Abstract STAP algorithms have been shown to be very effective in localizing and nulling ground clutter in homogeneous environments. However, false alarms due to moderate and strong clutter discretes continue to be a problem. This is because the depth of the adaptive clutter null depends on the amount of representative clutter in the sample correlation matrix. Since clutter is often quite non-homogeneous (e.g., due to varying terrain), conventional training strategies may not select range gates containing enough clutter. This, in turn, may lead to weaker clutter nulls, clutter breakthrough, and ultimately false alarms. New techniques, such as Clutter Editing, can identify some of these false alarms during the detection stage. However, a complementary approach that works directly in the adaptive nulling stage would be desirable. This presentation shows a data-adaptive method for selecting the range gates used to form the sample correlation matrix. The technique, called Power Selected Training, results in deeper clutter nulls by choosing a subset of the most powerful range gates for training. Using the Mountaintop Data Library, we demonstrate the improved clutter nulling ability of Power Selected Training when compared with conventional training approaches. We also examine performance issues such as SINR and MDV, as well as computational aspects.
Direct comments and questions to: email@example.com
© MIT Lincoln Laboratory. All rights reserved.