Robustness analysis of a gradient indentification method for a nonlinear wiener system

The gradient identification of the linear filter part of a nonlinear Wiener system with unknown output non-linearity is investigated with respect to robustness in a deterministic sense. In order to estimate accurately the linear filter coefficients, an adaptive nonlinear filter is placed at the output of the Wiener system which compensates the output non-linearity. Therefore, two adaptive algorithms work simultaneously. Local and global passivity relations are derived from which information on the robustness of the algorithm can be extracted