(An) approach for nonminimum phase system identification using third- and fourth-order cumulants

Abstract

In this thesis we consider the problem of estimating the parameters of moving average systems from the cumulants of noisy observations. The system is driven by an independent and identically distributed non-Gaussian signal which has an asymmetric probability density function. The measurement Gaussian noise may be either white or colored. We present an algorithm for identification of linear, time-invariant, nonminimum phase systems using the third- and fourth-order cumulants. Since the proposed algorithm does not use autocorrelation and makes use of higher-order statistics, it is blind to Gaussian noise. In the proposed algorithm some linear equations are obtained and the least squares method is adopted to solve them. Simulation results show that the proposed algorithm is more useful than some linear methods when the characteristic of noise is not known or when the signal-to-noise ratio is low.