High performance spectral analysis via two-sided AR modeling

Summary form only given, as follows. A new method of AR spectrum estimation is presented. It is called two-sided autoregressive (TSAR) spectrum analysis because an interpolative or smoothing model is postulated, as opposed to the predictive models used in AR modeling. The matrix equations arising in the proposed estimation procedures exhibit a special structure. The exploitation of these structures leads to fast solutions that reduce the total number of computations by an order of magnitude compared with straightforward approaches. Special attention is directed to the constrained two-sided AR model. Simulation examples show the higher resolution capability of the proposed method when compared with the maximum-entropy method.