Digital extrapolation spectral analysis based on ARMA model

摘要

A digital extrapolation spectral analysis method based on an autoregressive moving average (ARMA) model is described. A time series is fitted by an ARMA model and the parameters by which the series can be linearly extrapolated are estimated. A discrete Fourier transformation is performed for the extrapolated data. Simulations indicate that this technique can achieve high resolution power and reduce computation by means of a fast Fourier transformation. In cases when the estimation of the order is not accurate enough or is assumed to be a little higher, achievement of moderately high resolution confirms the robustness of this method.