Handling nonnegative constraints in spectral estimation

摘要: We consider convex optimization problems with the constraint that variables form a finite autocorrelation sequence, or equivalently, corresponding power spectral density is nonnegative. This often approximated by sampling density, which results in set of linear inequalities. It can also be cast as matrix inequality via positive-real lemma. The formulation exact, and solved using interior-point methods for semidefinite programming. However, these require O(n/sup 6/) floating point operations per iteration, if general-purpose implementation used. introduce much more efficient method complexity 3/) FLOPS iteration.