Understanding Several Power Spectrum Density Estimation Algorithms Based on the Problems Solved

Abstract

Starting from the estimation criterion and the essential problem of algorithmic computing, this paper introduces four kinds of spectral estimation methods commonly used in teaching: Periodogram to fit frequency vector under least square criterion，AR model estimation converts spectrum estimation into parameter estimation, MVDR converts spectrum estimation into filter weight coefficient estimation, and MUSIC converts spectral estimation into subspace estimation.Then we analyze and solve the equation, derive the assumption of signal and the defects caused by this assumption.