Robust linear prediction of band-limited signals
Henry M. Dante · 2003
The problem of getting a set of robust prediction coefficients which are independent of the underlying signal is considered. Explicit coefficients have been determined for when the sampling frequency is twice or three times the Nyquist sampling rate. It has been recently shown that it is possible to predict a bandlimited signal in terms of the past samples only. The authors show one set of such prediction coefficients which are easy to compute and show that as the number of past samples considered goes to infinity, the prediction coefficients which are easy to compute and show that as the number of past samples considered goes to infinity, the prediction error goes to zero.>