A Modified Burg Algorithm Equivalent In Results to Levinson Algorithm

Abstract

We present a new modified Burg-Like algorithm for spectral estimation and adaptive signal processing, that yield the same prediction coefficients given by the Levinson algorithm for the solution of the normal equations. An equivalency proof is given for both the 1D signal and 2D signal cases. Numerical simulations illustrate the improved accuracy and stability in spectral power amplitude and localization; especially in the cases of low signal to noise ratio, and (or) augmenting the used prediction coefficients number for a relatively short data records. Also our simulations illustrate that for relatively short data records the unmodified version of Burg Algorithm fail to minimize the mean square residual error beyond certain Order, while the new algorithm continue the minimization with Order elevation.