A modified direct whitening method for ARMA model parameter estimation

摘要

A modified version of direct whitening method which is called MDWM is proposed as an ARMA model parameter estimation technique. The direct whitening method(DWM) proposed by the author gives the parameter estimates which make the prediction errors uncorrelated. At some model, the implemented DWM algorithm falls at local minimum and give missing parameter estimates. This is similar in the well known PEM. To overcome this problem, MDWM which choose the consistent estimates among a large number of DWM estimates is proposed. In the simulation for the models which have poles and zeros are very close, the proposed method gives good estimates.