Non-Gaussian correlated process sampling for Bayesian cognitive target classification

Abstract

A cognitive signal processing system (for example, in radar or sonar) is one that observes and learns from the environment; then uses a dynamic closed-loop feedback mechanism to adapt the illumination waveform so as to provide system performance improvements over traditional systems. Current cognitive radar algorithms are designed only for target impulse responses that are Gaussian distributed to achieve mathematical tractability. Our research generalizes the cognitive radar target classifier to deal effectively with arbitrary non-Gaussian distributed target responses. Given exemplars of target impulse responses, our Bayesian illumination waveform design algorithm requires the ability to draw complex correlated samples from a target distribution specified by both an arbitrary desired probability density function and a desired power spectral density. This capability is realized using kernel density estimation and an extension of a new simple and efficient nonlinear sampling algorithm by Nichols et al. Simulations using non-Gaussian target impulse response waveforms demonstrate very effective target classification performance. We discuss practical issues with the application of the algorithms to real-world problems.