Nonlinear phase estimators based on the Kullback distance

This paper considers the design of phase estimators by combining concepts of stochastic nonlinear filtering and information theory. To propagate the involved probability density functions, adequate finite representations are needed. This is accomplished in this work by adopting minimum Kullback (1978) distance criteria. Applied to the important and paradigmatic cyclic phase estimation problem, our approach leads to consistent and systematic design methods. The resulting simple and parallelizable structure outperforms the commonly used extended Kalman-Bucy filter in tracking and acquisition situations. These features make the developed nonlinear filter suited to digital communications (carrier synchronization)