Simple, effective MA and ARMA techniques

An approach to MA estimation is proposed which uses a high-order AR estimate as an intermediate step. Although the AR estimate cannot directly represent deep spectral valleys, it is shown that much information is still contained in the AR coefficient sequence and can be extracted. Two methods of ARMA estimation are developed utilizing the MA technique. One estimates the poles by a previously known method and then estimates the zeros from the residual. The other first estimates the zeros, and then the poles from residual. The second method can also be used to extract the signal component of a high-order AR model after singular-value decomposition. Examples are given showing the techniques to be quite effective.