ASYMPTOTIC NOISE GAIN OF POLYNOMIAL PREDICTORS

Finite impulse response filters for the prediction of polynomial signa ls are considered. An expression for the asymptotic noise gain (as the filter length increases without bound) is derived. It is shown that t he asymptotic noise gain only depends on the polynomial order - in par ticular, it is independent of the prediction horizon. It is also shown that the noise gain forms a non-increasing sequence for increasing fi lter lengths. Numerical results that lend support to the theoretical f indings are included. (C) 1997 Published by Elsevier Science B.V.