Two semiparametric algorithms to recover a nonlinear characteristic in a Hammerstein
system are proposed. Both are obtained by incorporating a parametric component into the
kernel nonparametric algorithm. For small number of observations, their identification errors
are smaller than that of the purely nonparametric algorithm. The same idea is also proposed
for identification of linear dynamic component. Parametric instrumental variables estimate is
elastically substituted by the nonparametric correlation-based method, when the number of …