Skip to Main Content
In a recent correspondence Tzannes et al.  introduced a new form of minimum cross-entropy (MCE) spectral analysis based on the observation that normalized spectra and symmetric probability density functions (pdf's) are axiomatically indistinguishable. Thus, the form of the minimum cross-entropy posterior pdf can be used as a model for spectral estimation. In this correspondence we derive the same result without using a Lagrange multiplier methodology. We also establish the connection of this estimate to the cepstral representation of the signal and propose an efficient algorithm not requiring explicit solution of nonlinear equations.