Distortion compensation of nonlinear systems based on indirect learning architecture

Distortion compensation of nonlinear systems is an important topic in many practical applications. This paper concerns with linearization of nonlinear systems which can be modeled using Volterra series by connecting two adaptive nonlinear Volterra filters. The first one is a training filter connected in parallel with the nonlinear system and its kernels are estimated recursively. The second adaptive filter is a predistorter connected tandemly with the nonlinear system and its kernels are a copy from the training filter. Three recursive algorithms, namely: the recursive least squares (RLS), the Kalman filter (KF), and the recursive prediction error method (RPEM) algorithms, are developed and studied using numerical simulations. Simulation studies for time-invariant and time-varying nonlinear systems show that the KF and RPEM algorithms provide lower nonlinear distortion as compared to the RLS algorithm.