A Probabilistic Interpretation of the KL Spectrum

  • Seongbaek Yi (Division of Mathematical Sciences, Pukyong National University) ;
  • Park, Byoung-Seon (Deaprtment of Applied Statistics, Yonsei University)
  • Published : 2000.03.01

Abstract

A spectrum minimizing the frequency-domain Kullback-Leibler information number has been proposed and used to modify a spectrum estimate. Some numerical examples have illustrated the KL spectrum estimate is superior to the initial estimate, i.e., the autocovariances obtained by the inverse Fourier transformation of the KL spectrum estimate are closer to the sample autocovariances of the given observations than those of the initial spectrum estimate. Also, it has been shown that a Gaussian autoregressive process associated with the KL spectrum is the closest in the timedomain Kullback-Leibler sense to a Gaussian white noise process subject to given autocovariance constraints. In this paper a corresponding conditional probability theorem is presented, which gives another rationale to the KL spectrum.

Keywords

References

  1. IEEE Trans. Acoustics v.38 An information theoretic spectral density Choi;B. S.
  2. IEEE Trans. v.39 A Time-domain interpretation of the KL spectrum Choi;B. S.
  3. IEEE Trans. Information Theory v.33 Conditional Limit theorems under Markov conditoning Csiszar;I.;Cover;T. M.;Choi;B. S.
  4. 3rd ed. Multivariate Statistical Methods MOrrison;D. F.
  5. Prob. Th. Rel. Fields v.72 Berry-Esseen theorems for quadratic forms of Gaussian Stationary Processes Taniguchi;M.
  6. IEEE Trans. Information Theory v.27 Maxmum entropy and conditional probability Van Campenhout;J. M.;Cover;T. M.