ARMA processes in multirate filter banks with applications to radar signal classification

Abstract

Considers stochastic modeling in a scale space defined by multirate filter banks using autoregressive moving average (ARMA) models. The authors show that signals at coarser scales in analysis filter banks follow ARMA models if the signal at a finer scale is an ARMA process. The model for a coarser scale signals can be identified from the model of a finer scale signal. Reconstruction of a finer scale signal from a partial set of decomposed signals at a coarser scale is considered as an optimal estimation problem. The authors developed a recursive minimum mean square error (MMSE) estimation algorithm for reconstruction of a finer scale signal. The stochastic modeling approach in scale space is applied to classification of radar signals. >