Multiple-Model Based Particle Filters for Frequency Tracking in $\alpha$-Stable Noise

In the work presented here, we contribute to two extensions to the sequential Monte Carlo (SMC) or particle filtering approach for on-line sinusoidal frequency tracking in. First, we propose to employ multiple dynamical models for better parameter tracking at different times in the light of observations as the unknown frequency and its rate of change are time varying. Second, α-stable noise, which is a generalization of the traditional Gaussian process, is considered as the additive disturbance in the observed signal. In other words, the proposed method is required to estimate both the unknown time-varying frequency and the characteristics of the α-stable process. According to computer simulation results, it is found that the proposed multiple-model method performs better in terms of smaller estimation error than the case when only a particular single model is employed.