Blind identifiability of quadratic non-linear systems in higher-order statistics domain

Quadratic non-linear systems are widely used in various engineering fields such as signal processing, system filtering, predicting and identification. Some conditions to blindly estimate kernels of any discrete and finite ext ent quadratic system in the higher-order cumulants domain are introduced in this paper. The input signal is assumed as an unobservable i.i.d. random s equence which is viable for engineering practice. Due to properties of the output third-order cumulant functions, identifiability of the non-linear sy stem holds even if the system's output measurement is corrupted by a Gaussi an random disturbance. It provides a useful starting point for implementati ng the identification of a truncated Volterra non-linear system using conve ntional techniques or neural network methodologies. (C) 1998 John Wiley & S ons, Ltd.