System identification using higher order statistics

This paper presents a novel approach for the identification of unknown systems, using measurements of the output signal only. It describes a convergent adaptive algorithm that identifies the parameters of the unknown system whether a minimum phase or non-minimum phase one. The identification process is achieved through exciting the adaptive system by an independent random identically distributed signal i.i.d., and minimizing-in a least squares sense-the difference between the cumulants of the desired response and the output of the adaptive system. In the general ARMA process, the adaptive system is modeled as discrete orthogonal sections. Illustrative examples are given to show that the proposed method manages to identify unknown systems that known published fail to identify. The identification is shown to be successful even when the desired signal is contaminated with noise.