Estimating of Volterra kernels in case of moment hierarchy and orthogonal representation (Q679257)

Two distinct approaches of Wiener [\textit{N. Wiener}, Response of a nonlinear device to noise. MIT Radiation Laboratory Report 165 (1942), and Nonlinear problems in random theory, John Wiley, New York (1958; Zbl 0121.12302)], developed to identify nonlinear systems are considered. Due to the fact that an excessive number of coefficients are needed to characterize the system of interest, Wiener's approaches are impractical and difficult to implement on computers. To resolve this problem, the method of the orthogonalization of the input data in the Gaussian case and the method of moment hierarchy are established and considered. When the expansion contains orthogonal terms, the Volterra kernel estimation is given. A relationship between the kernels in case of orthogonal and nonorthogonal expansion is established. At the end, numerical examples are employed to compare the two methods established in this work.