Linear prediction, maximum flatness, maximum entropy, and AR polyspectral estimation

The authors present a theoretical foundation for polyspectral estimation and modeling of non-Gaussian autoregressive (AR) processes which includes a new higher-order-statistics (HOS)-based linear prediction error filter and associated linear prediction polyspectral estimator, and a maximum higher order entropy polyspectral estimator, and considers the equivalences among these polyspectral estimators.<
>