Unbiased two-windows approach for Welch's method

TL;DR

This paper introduces a new two-window method for Welch's spectral estimation that is unbiased for signals with bounded correlation lengths, improving accuracy over traditional approaches especially for complex signals.

Contribution

The paper proposes a novel two-window approach for Welch's method that eliminates bias for signals with bounded correlation lengths, unlike traditional methods.

Findings

The new approach is unbiased for signals with bounded correlation lengths.

Numerical experiments demonstrate the method's advantages over traditional Welch's method.

Traditional Welch's method is biased for finite windows on complex signals.

Abstract

Periodogram methods are widely used for the estimation of power- and cross-spectra, of which Welch's method is the most popular. Previous studies have analyzed the variance of the power spectra estimates and developed analytical probability functions, showing that the approach is unbiased when applied to white-noise signals or in the limit of infinite window lengths. However, no explicit expression for the estimation bias is available for more complex signals when finite windows are used. In this study, we show that, for finite window lengths, Welch's method is biased for all signals other than the white-noise signal. A novel two-window approach that is unbiased when applied to signals with bounded correlation lengths is proposed. Numerical experiments are used to illustrate the advantages of the novel approach.