Parameter estimation of Hammerstein model based on a gradient algorithm inwavelet domain

For the discrete nonlinear Hammerstein model with the noise corrupted output data, a gradient method is proposed to estimate the parameters of the model with the input-output data in wavelet domain directly. With the wavelet theory being developed it plays an important role in signal processing. By means of wavelet transform, the signal has both characteristics of time and frequency and becomes a signal in wavelet domain, and increasing the ratio of signal to noise. The parameters of model are estimated by the wavelet gradient method, compare with the gradient method in time domain, the proposed method is effective by the simulation.