Use of the most significant autocorrelation lags in iterative ME spectral estimation

Many methods of spectral analysis are based either directly or indirectly on a set of autocorrelation values estimated from the available data. A good selection of autocorrelation Lags can improve the quality of the spectral estimate at a given computational cost. To demonstrate the above possibility, this paper shows how to select Lags corresponding to the most significant values of the autocorrelation. In this way, one obtains better estimates than those found using the standard method, namely, the technique proposed by Lim and Malik [1] for (iterative) ME spectral analysis. Several examples arc considered to illustrate this possibility.