An exponentially convergent adaptive algorithm for time-varying IIR filters

Conditions for exponential stability are established for a class of adaptive algorithms that estimate the parameters of a time-varying linear filter using a basis function approach. These algorithms are useful in situations calling for adaptive filters when the ideal filters vary at rates that cannot be considered slow, such as in the study of renal hemodynamics and for some mobile radio channels. Exponential convergence is established by extending a Lyapunov stability analysis for time-invariant adaptive IIR filters to the time-varying case.