MULTILAYER VOLTERRA FILTER AND ITS APPLICATIONS

The paper presents the application of the successive linearisation to the neural network implementation of the Volterra filter. Applying the signal flow graph approach, the new learning rules for adaptation of weights of the multilayer network structure obtained are given. The mu ltilayer structure presented is applied to signal processing, includin g identification of the parameters of the plant, noise cancelling and signal prediction. The results of the simulation of the filter in thes e three basic application modes are presented and discussed.