Numerically efficient mean squared error threshold SNR prediction for adaptive arrays

Abstract

The method of interval estimation (MIE) is an established technique for extending asymptotic mean squared error (MSE) predictions like the Cramer-Rao bound to lower signal-to-noise ratio. While application of MIE to the adaptive array problem was successful in [1], the numerical integration required to compute the pairwise error probabilities central to MIE is computationally expensive. This is primarily due to the double integral required, moreover, the integrand itself involves the Marcum Q-function, a specialize function that can be represented as an integral or infinite series. System analysis and design often requires computing MSE performance over a wide search space that easily demands hundreds to tens of thousands of repeated calculations of the pairwise error probabilities. To support this demand two approaches to approximating the required error probabilities are explored herein, one yielding a near ~235 times speedup factor in computation without major loss in accuracy of MSE prediction.