Neural network identification of digital satellite channels: the adaptive nonlinear enhancer

Abstract

This paper presents a neural network approach for modelling satellite communication channels which are equipped with nonlinear devices (travelling wave tubes (TWT)). The model is based upon the use of multilayer neural networks (MLNN) which are trained with the back propagation (BP) algorithm. It is shown from simulation results that the TWT characteristic can be extracted by the neural net model, as well as the linear filters included in the channel. The learning process is performed by using the channel input and output signals. The neural network model works as an adaptive nonlinear enhancer.