A REDUCED PARAMETER SECOND-ORDER VOLTERRA FILTER WITH APPLICATION TO NONLINEAR ADAPTIVE PREDICTION OF CHAOTIC TIME SERIES

Abstract

A reduced parameter second-order Volterra filter (RPSOVF) which is constructed by the multiplication-coupled two linear FIR filters, and its nonlinear normalized least mean square (NLMS) algorithm is proposed; and this RPSOVF with nonlinear NLMS algorithm are used to make adaptive predictions of chaotic time series. The rule of selecting convergent assistant parameters of the nonlinear NLMS algorithm is obtained. Experimental results show that this reduced parameter second-order Volterra filter with the nonlinear NLMS algorithm can be successfully used to make adaptive predictions of chaotic time series, and the modified nonlinear NLMS algorithm enables RPSOVF to converge and stabilize.