The quadratic extended Kalman filter

Abstract

This paper extends the extended Kalman filter (EKF) to cover those scenarios in which the input signal is unknown and quadratic processing is required to estimate and track some parameters of interest. The optimal quadratic tracker is found to incorporate the fourth-order statistical knowledge about the excitation signal. To illustrate these conclusions, we have studied the problem of direction-of-arrival (DOA) tracking of digitally-modulated signals. Simulations have reflected the advantage of considering this information in terms of acquisition and steady-state performance.