Simultaneous Detection and Estimation for Diffusion Type Signals

Kyu San Han

Doctoral Dissertation, Year: 1986, Advisor: John S. Baras

Abstract

We consider a problem of combined detection and estimation when signals corresponding to M-ary hypotheses can be represented as outputs of M distinct parameterized stochastic dynamical systems of the Ito type. For a general synthesis of detector and estimator we take a jointly Bayesian approach to the combined detection and estimation under a suitable class of joint cost functions that combine a cost of misdetection with a cost of estimation error, and then utilizing the nonlinear filtering theory we derive subsequent jointly Bayesian receiver structures.

Secondly, based on the Chernoff bounds we analyze detection performances for both full and partial observation of signals generated by several kinds of finite dimensional nonlinear stochastic dynamical systems. Also performance evaluation in the parameter estimations is treated from a detection point of view.