Robust Space-Time Adaptive Processing Using Projection Statistics

In this paper, projection statistics (PS) are applied to detect and mitigate outliers in adaptive processing algorithms. Outliers present in training data result in slow adaptive processor convergence. In radar, these outliers may arise from clutter discretes, desired targets, or jammers. PS provide a computationally tractable technique for identifying and mitigating outlier data samples prior to adaptive processing. We demonstrate that well known processing methods, such as sample matrix inversion and its variants can be made robust to these outlier impairments by incorporating PS into the algorithm formulation.