Sequential MCMC Estimation of Nonlinear Instantaneous Frequency

Abstract

Instantaneous frequency (IF) estimation of signals with nonlinear phase is challenging, especially for online processing. In this paper, we propose IF estimation using sequential Bayesian techniques, by combining the particle filtering method with the Markov chain Monte Carlo (MCMC) method. Using this approach, a nonlinear IF of unknown closed form is approximated as a linear combination of the IFs of non-overlapping waveforms with polynomial phase. Simultaneously applying parameter estimation and model selection, the new technique is extended to the IF estimation of multicomponent signals. Using simulations, the performance of this sequential MCMC approach is demonstrated and compared with an existing IF estimation technique using the Wigner distribution.