Detection of spectrally equivalent parametric processes using higher order statistics

The paper addresses the problem of detecting two spectrally equivalent parametric processes (SEPP): the noisy AR process and the ARMA process. Higher-order statistics (HOS) are shown to be effective for detection. Two HOS based detectors are derived and compared. The fist detector studies the singularity of a HOS-based Yule-Walker matrix. The second detector filters the data by an AR filter estimated from the data; the residual HOS are then shown to be effective for the SEPP detection problem