Neural network identification and characterization of digital satellite channels: application to fault detection

The paper proposes a neural network technique to adaptively model and characterize digital satellite channels. The neural network model allows to identify each component of the channel by the use of the channel input-output signals as learning data. This technique was applied to fault detection in digital satellite links, especially those arising in on-board devices. The paper gives simulation examples of changes in the on-board filter characteristics. Our adaptive method allows to determine the origins of the changes and gives the new characteristics of the channel.