An adaptive technique for designing minimum phase models

The authors have developed a novel filter structure, called a stochastic filter, that consists of a bank of fixed filters with a set of corresponding probabilities. The fixed filters can be viewed as the basis set of filters for the stochastic filter, with the probabilities determining the specific realization represented by the stochastic filter. If the probabilities are allowed to vary with time, the stochastic filter is an implementation of an adaptive filter. The authors present a configuration, using FIR (finite impulse response) fixed filters in the stochastic filter, that can be used to adaptively model unknown systems. It is shown theoretically that this form of adaptive stochastic filter converges to a minimum phase FIR filter model