ABSTRACT
The autoregressive (AR) model plays a key role in modern spectral analysis. As
a parametric alternative to the nonparametric periodogram, the AR model is a
linear dynamic system in the form of a difference equation with a few unknown
parameters. By estimating these parameters and examining the transfer func-
tion of the resulting dynamic system, one is able to accomplish the objectives of
spectral analysis. In this chapter, we discuss two ways of using the AR model to
estimate the signal frequencies in a time series with mixed spectrum. One way
is to fit an AR model to the time series and identify the peaks of the resulting AR
spectrum like a periodogram. Another way is to reparameterize the sinusoidal
signal by an AR model and estimate the corresponding AR parameters instead.