Covarianceestimationforhighdimensionalvectorsisaclassicallydifcultprobleminstatisticalanalysisandmachinelearning.Inthispaper,weproposeamaximumlikelihood(ML)approachtocovarianceestimation,whichemploysanovelsparsityconstraint.Morespecically,thecovarianceisconstrainedtohaveaneigendecompos