Demodulation of noisy phase or frequency modulated signals with Kalman filters

The paper addresses the problem of estimating the argument of a phase or frequency modulated sinusoidal signal of unknown amplitude in the presence of additive Gaussian white noise of known variance. It is assumed that a state space model for both the unknown phase and the unknown amplitude is available either as an exact or approximate description of the reality. The problem is solved by an extended linearised Kalman filter. The basic idea of using extended Kalman filters was previously described but has not been used to perform simultaneous amplitude and phase demodulation.<>