Semiparametric identification of Hammerstein systems using input reconstruction and a single harmonic input

We present a two-step method for identifying SISO Hammerstein systems. First, using a persistent input with retrospective cost optimization, we estimate a parametric model of the linear system. Next, we pass a single harmonic signal through the system. We use l-delay input reconstruction with the parametric model of the linear system to estimate the inaccessible intermediate signal. Using the estimate of the intermediate signal we estimate a nonparametric model of the static nonlinearity, which is assumed to be only piecewise continuous. This method is demonstrated on several numerical and experimental examples of increasing complexity.