In this paper, an adaptive identification method is presented, it is developed for discrete Hammerstein systems. We adopt a frequency-domain technique by using sampled input–output data. The excitation signals are chosen to be the fundamental harmonics, by which one could get approximating estimate of frequency responses to linear subsystem. By using adaptive rational orthogonal basis functions, we obtain approximations of linear subsystem, it is carried out through selecting poles of the basis functions under some criterion. Meanwhile, efficient estimates of both the nonlinear part and linear part could be obtained in the presence of noise.