Study of model-order selection algorithms and their applicability to adaptive clutter suppression

The problem of predicting the optimum autoregressive (AR) model order is addressed in the context of adaptive radar clutter suppression. The applicability of existing model-order selection criteria to AR clutter models was studied through a computer simulation of a scanning Doppler radar. Reasonably good predictions were achieved with the Akaike final-prediction-error criterion and the Parzen criterion-autoregressive-transfer when the coefficients obtained in one scan of the radar antenna were used on the data obtained in the next scan.<
>