Estimación espectral de ondas oceánicas por máxima entropía

Abstract

This paper discusses the statistics of four criteria to estimate the order of the autorregressive process (AR) used to fit ocean wave data. These criteria are:
a) Final Prediction Error (FPE)
b) Autorregressive Transfer Function (CAT)
c) slope change of FPE function
d) Partial Autocorrelation (PA)
The Maximum Entropy Method (MEM TECHNIQUE) is used to estimate the spectra of two sets of data: with and without swell energy contributions. Each time series is 17 minutes long. It was found that the optimum filter prediction length is between 20 and 30 for signals without swell energy contribution, and between 35 and 45 otherwise.
MEM estimates are compared with conventional ones (smoothed periodogram). From MEM spectra estimates of different AR process order, the spectral moments are computed up to order four and no significant differences are found.