A novel architecture to model non-linear systems

Abstract

This paper shows a new architecture specially thought to model non-linear systems (NLSs). At first, it was applied only to memoryless systems but then it developed to solve a more general problem, NLSs with memory. The result is a new filter, based on the Fourier transform, that the authors have named “K-filter”. Important features of the K-filter are its nonlinear behaviour and second, that it profits from a temporal diversity of the input signal in order to provide itself with memory. At the end of the paper, the K-filter is used to solve an identification problem of a communication system which behaves nonlinearly due to the response of the amplifiers and which also has memory introduced basically by the channel response. The simulation results will provide an evaluation of the K-filter.