A NFxLMS algorithm with Initial Subsystem Estimates for digital predistortion of Wiener systems

Adaptive predistortion of nonlinear systems described as IIR Wiener models is discussed in this paper. The predistorter is modeled as an IIR Hammerstein system. The parameters of the linear and nonlinear blocks of the Hammerstein predistorter are estimated simultaneously using the Nonlinear Filtered-x Least Mean Squares (NFxLMS) algorithm. In the NFxLMS algorithm, the nonlinear system is assumed to be known or accurately identified. In this paper, a novel NFxLMS with Initial Subsystem Estimates (NFxLMS-ISE) algorithm is developed to avoid accurate identification of the nonlinear system. Simulation study shows that the NFxLMS-ISE algorithm can efficiently precompensate the nonlinear distortion similarly as the NFxLMS algorithm assuming that the nonlinear system is known.