Linear in the parameters identification for classes of systems

System identification is often required in situations in which the unknown system is drawn from a class of systems whose characteristics are imprecisely specified. In this paper, a method for selecting an identifier model set is developed which matches the model set to the class of unknown systems. A priori information in the form of impulse responses for examples taken from the system class is used in choosing the model set. The model set is constrained to have a linear in the parameters structure, so that online identification algorithms with robust convergence properties are applicable to the problem. This approach is shown to produce an identification procedure tailored to the class of systems in hand, with the benefit of providing reasonably accurate system models while using a limited number of adjustable model parameters