Tracking of non-stationary signals with periodic instantaneous frequencies using extended Kalman filters

In situations when the instantaneous frequency (IF) of a non-stationary signal is periodic, estimation of signal parameters by continuous adaptation is expected to provide superior performance to that by block processing approaches as a consequence of the ability to accumulate information over time. The extended Kalman filter is applied to exploit the periodicity of the IF for the tracking of frequency modulation (FM) signals in both the time and frequency domain. No prior knowledge of the IF model is required, and an efficient algorithm for the initialization of filter conditions is presented. Simulation results reveal the outstanding performance of the proposed method in tracking nonstationary signals with IF of different model orders