Abstract:

To solve the problem of analysis and computation during using a finite impulse response (FIR) model with high orders, orthonormal basis functions are constructed by utilizing partial system information and then introduced into traditional FIR modeling. A generalized FIR modeling method is proposed. This method prefilters input data by the orthonormal basis functions in order to get some dynamic characteristics. Unfortunately, the gradient algorithm could not be directly used in parameter identification because the prefiltered data do not have time shift characteristic anymore. Combined with the principle of recursive least mean squares, the parameter calculation formulas are presented. Lastly a numerical example is studied. The results shows that the generalized FIR modeling method has good performances and can fully satisfy the requirement of system identification.