Abstract
Recently, two types of modifications are proposed over the Burg method to reduce frequency estimation errors for short data records of the sinusoidal signals. The first modification uses an optimum-tapered window function for weighting the linear prediction error energies in the minimization process. The second modification uses the measured energy of data samples for weighting. In the present paper, we propose to use both the tapering and energy-weighting and show that the resulting optimum-tapered energy-weighted Burg method reduces the frequency estimation errors for sinusoids further. This method is studied for a number of examples of sinusoidal signals and its performance is discussed with respect to line splitting, frequency bias, variance and root-mean-square error.
Indexing terms: