The Cable and Telecommunications Professionals' Reference, 3rd Edition

18  Particle Filtering in Digital Communications

E. Punskaya, A. Doucet, and W. J. Fitzgerald

Introduction

In many engineering applications, one needs to extract the signal from the data corrupted by additive random noise and interference of different kinds in order to recover the unknown quantities of interest. This is a problem of scientific inference, and to carry out consistent reasoning and inference, one may use the Bayesian paradigm.

The main challenge of the Bayesian approach, however, is the fact that quite often it requires multidimensional integration over parameter space, which is generally only possible when the models are linear in the parameters and when one makes a simplifying noise assumption, such as Gaussianity. In more realistic ...