In conventional one-step forward linear prediction, an estimate for the cur
rent sample value is formed as a linear combination of previous sample valu
es. In this paper, a generalized form of this scheme is studied. Here, the
prediction is not based simply on the previous sample values but to the sig
nal history as seen through an arbitrary filterbank. It is shown in the pap
er how the coefficients of a modified model can be obtained and how the inv
erse and synthesis filters can be implemented. Various properties of such s
ystems are derived in this article. As an example, a novel linear predictiv
e system using inherently logarithmic frequency representation is introduce
d.