The previous chapters surveyed the general theory. In this chapter we will specialize the analysis to frequency and spectrum estimates. Our ultimate aim is to derive explicit Bayesian estimates of the power spectrum and other parameters when multiple nonstationary frequencies are present. We will do this by proceeding through several stages beginning with the simplest spectrum estimation problem. We do this because as was shown by Jaynes  when multiple well-separated frequencies are present [|ω j - ω k | ≫ 2π/N], the spectrum estimation problem essentially separates into independent single-frequency problems. It is only when multiple frequencies are close together that we will need to use more general models.
KeywordsPower Spectral Density Discrete Fourier Transform Accuracy Estimate Single Frequency True Frequency
Unable to display preview. Download preview PDF.