Cepstrum constraints in ME spectral estimation

Current trends in spectral estimation techniques follow basicaly two guidelines; by introducing modifications in the objetive function to be minimized or, keeping the objetive, by modifying the set of constraints in the minimization procedure. This work can be encompassed in the second sense, showing the potential of cepstrum constraints in spectral estimation methods. The special features associated with this constraints allow to use raw approximations in order to linearize or simplify the computations involved in the procedure maintaning within margins of adequate quality the resulting estimate. Being S(w) the spectral estimate of a random process {x} and given a data sample x(n) (n=O,M—1), an ME estimate have to minimize the objetive function (1) (7T log S (w) dw 7 j x in accordance with a set of constraints. The set of constraints use to be the autocorrelation estimate (2), which is usually obtained from the lOFT of the periodogram 5(w) corresponding with the data sample Ii I