An information theoretic spectral density
B. S. Choi · IEEE Transactions on Acoustics Speech and Signal Processing · 1990
An information theoretic spectrum which minimizes the Kullback-Leibler (1951) information number subject to the first p+1 autocovariance terms is proposed. The KL spectrum includes the maximum-entropy spectrum and the ARMA (autoregressive moving-average) spectrum as special cases. A method is proposed for modifying a spectral estimate, on the basis of the results for the KL spectrum, so that the revised spectral estimate is more loyal to the given observations than the primary estimate.>