In this paper, a procedure to identify the weighting function of a system is presented. A new type of criterion is employed. It consists of the term concerning the a priori information of the weighting function and the ordinary squares integral term of the response error. It is shown that the optimal weighting function under the proposed criterion function can be determined easily by solving an optimal tracking problem, if the output of a linear free dynamical system is supplied as an input to the system to be identified.
The proposed criterion function is also interpreted from the viewpoint of estimation theory and found to be the same as the maximum likelihood function under the proper conditions.
A computational algorithm of the procedure to identify the weighting function of a single input and single output system is presented. It is also shown that the algorithm can be easily extended for multivariable systems.
Numerical computations of several examples show that the proposed procedure is satisfactory even when the system is observed only for a short period.