Advances in Lee–Schetzen Method for Volterra Filter Identification

Improving the approximation ability of Volterra series identified with a cross-correlation method

“…In [12], the same authors show how input nonidealities can affect kernel estimation. In particular, explicit expression of residuals up to the sixth order are given.…”

“…The effect of input amplitude on convergence was investigated in [12], where it has been shown that identification input non-idealities makes the output MSE a function of the input variance.…”

Multivariance nonlinear system identification using Wiener basis functions and perfect sequences

“…The corrections needed in diagonal points identification require to recompute the lower odd/even order kernels for each odd/even order kernel to be identified [7]. In contrast, exploiting the Wiener basis functions and the WN filter in (3), the kernels k l,... become independent of each other and can be separately estimated using (4).…”

“…Moreover, an exact white Gaussian input cannot be generated due to the limitation of the input signal length and to the input amplitude saturation. Furthermore, the central moments of a Gaussian input soon depart from ideal values as the moment order increases unless millions of values are used [4]. Some improvements of the first implementations of the crosscorrelation method (e.g., Lee-Schetzen [1]) were provided in [4], [5] to reduce the input non-ideality and errors due to model order truncation that affect the kernels diagonal points [4].…”

“…The non-idealities of input noise make the output mean square error (MSE) a function of the input variance [4]. In particular, an accurate This work was supported in part by DII Research Grant and by DiSPeA Research Grant.…”