GLOBALLY CONVERGENT ADAPTIVE IIR FILTERS BASED ON FIXED POLE LOCATIONS

A new class of adaptive filters, dubbed fixed pole adaptive filters (F PAF's), is introduced, These adaptive filters have infinite impulse re sponses, yet their adaptation exhibits provable global convergence, Go od filter performance with a relatively small number of adapted parame ters is permitted by the new filter structure, thus reducing the compu tational burden needed to implement adaptive filters, The implementati on and computational complexity of the FPAF is described, and its mode ling capabilities are determined, Excitation conditions on the biter i nput are established that guarantee global convergence of a standard s et of adaptive algorithms, Some methods are described for selecting th e fixed pole locations based on a priori information regarding the ope rating environment of the adaptive filter, The FPAF is tailored to app lications by such a procedure, enabling improved performance, In examp les, the FPAF is shown to achieve a smaller minimum mean square error, given an equal number of adapted parameters, in comparison with adapt ive FIR filters and adaptive filters based on Laguerre and Kautz model s.