The Monte Carlo framework for adaptive IIR filter & ARMA parameters estimation

Adaptive filter is widely used in many cases and usually formed in FIR structure. However, in multi-path wireless channel equalization, the channel can be presented as full zero-point model, and the difficulty will be encountered for equalization using adaptive FIR filter. In this situation, the channel can be compensated with satisfy characteristics by using adaptive IIR filter and obtain better performance than FIR. Unfortunately, the general IIR algorithm is difficult to perform filtering with robust and downright. In another signal processing field, such as spectrum estimation and system identification, the classical approach is to model the signal (or system) as ARMA process, i.e., a signal can be regarded as an output of a filter with some zero-pole in system function excited by white noise. Compared with AR model, in ARMA process, although there are some faulty algorithms before which bring troubles in high complexity and uncertain result, but the mature technology is absence as yet. Similarly, ARMA model can also be equivalent with a IIR filter, that´s to adjust the parameters to enable the filter´s output close with a reference signal. This paper proposes a new framework for adaptive IIR filter and ARMA parameter estimation based Monte Carlo particle filter, and the agility is demonstrated by computer simulation. The degeneracy phenomena will not occur in filtering. Simulation results indicated that the new method is feasible.