Recently, an optimal prefilter predictor for narrow-band lowpass signa
ls was introduced by Laakso and Ovaska. The prefilter predictor consis
ts of an application-specific fixed IIR filter cascaded with an analyt
ically optimized FIR postfilter. Using this approach, the FIR postfilt
er was designed such that polynomial signals of a given order are pred
icted to be unchanged, with additional degrees of freedom used for min
imization of the noise gain. In this paper, we enhance the original pr
efilter approach for more practical filter specifications. Instead of
inflexible polynomial constraints and full-band noise power minimizati
on, we define an explicit passband and stopband as in conventional fil
tering. The new analytical design method enables the flexible control
over the frequency domain characteristics of the predictor. Design exa
mples are provided to illustrate the indisputable benefits of the new
predictor class. (C) 1995 Academic Press, Inc.