Wald Sequential Detection with Non-Guassian Pulsed Radar Data Using the Zakai Equation

Abstract

The "Optimal" Wal sequential hypothesis test for diffusion signals is presented. The result is a threshold test with explicitly computable thresholds. Five possible schemes for a numerical implementation of the test are given. A comparison of the different implementations and analysis of the detectors performance is done for the radar problem of ship versus chaff target discrimination using lognormal and Rayleigh models respectively. Parameter estimation for the lognormal and Rayleigh cases is also studied. Finally, a signal estimation scheme is presented utilizing the conditional expectation of the signal computed from the conditional density of the underlying state, which is the solution to the Zakai equation.