Instantaneous Frequency Tracking for Polynomial Phase Signals Based on Extended Kalman Filter

Abstract

This work presents an effective approach for tracking the instantaneous frequency of polynomial phase signals. Compared with the conventional methods based on phase differentiation or time-frequency representation, the proposed method regards the polynomial phase signal as a state-space model. The polynomial phase is approximated by the local polynomial, and the corresponding coefficients constitute the state vector. Hence, the tracking of local polynomial phase coefficients is converted into a procedure of solving the state-space model. To determine the instantaneous phase, the extended Kalman filter is applied to solve the state-space model. Finally, the polynomial regression and the O’Shea refinement strategy are combined to improve the results to reach the Cramér-Rao lower bound. The computational complexity of our algorithm is O(K2•N). Simulation results indicate that the presented approach also preserves similar accuracy compared with the state-of-the-art.