Multitaper Spectral Estimation

Abstract

With discretely sampled andMultitaper method|( truncated data, a power spectrum computed simply by squaring and summing the real and imaginary components of the Fourier transform provides a biased estimate of the true power spectrum. And while windowing with a single taper decreases spectral leakage, it does not reduce the variance of the spectrum. This chapter introduces a class of multiple tapers called discrete prolate spheroidal sequences, which are designed to reduce leakage by solving the so-called concentration problem. Furthermore, they form an orthogonal set meaning that their spectrum estimates are independent, allowing for averaging and an associated reduction of variance.