Linear Prediction of Bandpass Signals Based on past Samples

This paper describes methods of computation of prediction coefficients for bandpass signals which provide highly accurate predictions from past samples of a signal based on the sampling rate, the bandwidth, and the center frequency of the signal. Two methods are described, one based on solving a matrix system and another on eigenvectors related to and extending the Discrete Prolate Spheroidal Sequences. In contrast to the standard LPC method, the resulting prediction coefficients are independent of the prediction time. Several examples are given.