On the optimal number of scales in estimation of fractal signals using wavelets and filter banks

In this paper we deal with the problem of finding the optimal number of scales used in a multiscale Wiener filter for obtaining the minimum mean-square estimation error of fractional Brownian motion (fBm) in noise. Several simulations are presented avoiding simplificative hypotheses previously used and considering also the effects of aliasing. Furthermore, it is shown that the mean-square error does not a strictly decreasing function with respect to the number of scales J. In all the analyzed cases, the optimal number of scales is J less than or equal to 6. (C) 1997 Elsevier science B.V.