Resolution capability of nonlinear spectral-estimation methods for short data lengths

The performance of nonlinear power spectral estimation methods is studied from the viewpoint of resolution. The methods considered are the maximum entropy method, the maximum likelihood method, the least-squares method, Prony´s energy spectral-density method, the singular-value decomposition technique and the autoregressive moving-average power spectral-estimation method. The data model used comprises two sinusoidal signals of equal amplitudes immersed in white Gaussian noise. The minimum resolution that can be obtained for all the power spectral-estimation methods considered has been studied for different data lengths and signal-to-noise ratios and the results are tabulated.