Recursive autoregressive spectral estimation by minimization of the free energy

A digital signal processing technique applicable to power spectrum estimation, designated as the minimum free energy method, is described. With no a priori model assumption and no attempt to extract special features such as sinusoids, one can obtain high resolution even with high noise contamination of the measured signal. The technique is demonstrated by modification of the Burg recursive method of spectral analysis. A recursive minimum free energy method in which the reflection coefficient is chosen at each step is proposed for minimizing the free energy (the Burg energy measure minus the product of a temperature and an information entropy). The method produces a spectral estimator more impervious to high noise contamination than the Burg method and diminishes the Burg tendency to produce spurious peaks.<
>