Non-Gaussian clutter modeling and application to radar target detection

The performance of a Gaussian receiver in non-Gaussian clutter is significantly inferior to that of an optimum receiver for the detection of weak targets. Knowledge of the clutter distribution is a key element to obtaining improved performance. This requires the specification of a suitable multivariate non-Gaussian probability density function (PDF) for the clutter. Spherically invariant random vectors (SIRVs) are used for the characterization of non-Gaussian clutter data. The Ozturk (Rangaswamy et al. 1993; Ozturk and Dudewicz 1992) algorithm, an efficient technique for approximating univariate distributions, is extended to the case of multivariate SIRV distributions. The performance of a novel adaptive non-Gaussian receiver, which utilizes the Ozturk algorithm, is evaluated for a K-distributed clutter example. This receiver shows significant improvement over the Gaussian receiver and closely matches the performance of the K-distributed receiver