Power spectrum estimation from values of noisy autocorrelations

Abstract

The problem of estimating the power spectrum from noisy autocorrelation values is considered in this paper, and it is proposed that in order to reduce errors, oversampling of the available time domain data should be employed. The oversampling problem is discussed from the frame theory point of view, and it is shown that the frame reconstruction represents an improvement upon the standard correlogram, windowing, and autoregressive modelling approaches.