Generation of rejection method bounds for spherically invariant random vectors

Based upon the central limit theorem, random clutter returns are commonly modeled as Gaussian. Nevertheless, many situations arise in practice where the data are clearly non-Gaussian, as is seen with "spiky" radar clutter. Spherically invariant random vectors (SIRVs) are especially attractive for modeling correlated non-Gaussian clutter. This paper discusses the computer simulation of SIRVs for Monte Carlo purposes using the rejection method. A key requirement of the rejection method is the ability to find a tight bound of the probability density function, from which random samples can be readily generated. An automated technique for generating this bound for the SIRV probability density function is presented.