AModified Power Spectrum Estimator for Minimizing the Prediction Error Energies

The power spectrum estimation plays a significant role in the marine monitoring and the quality of spectrum will seriously affect the results of marine monitoring. There have been various algorithms on power spectrum estimation, one of which is Burg algorithm. Though this method can effectively improve the resolution of spectral estimation, it will cause the shifting of peak frequency and the splitting of spectral lines accompanying spurious frequency components. In this paper, a novel method for minimizing the prediction error energies is presented, which is based on the optimum tapered Burg algorithm. The proposed method is implemented in two aspects. Firstly, the first order reflection coefficient is calculated by minimizing the averaged optimum tapered forward and backward prediction error energies of the third order filter. Then, the residual factor is introduced to estimate the model order. In the end of the paper, the experimental results demonstrate that in contrast to Burg algorithm and optimum tapered Burg algorithm, the proposed algorithm can not only effectively improve the spectrum resolution, lower the degree of peak frequency shifting and spectrum lines splitting, but also be able to alleviate the generation of spurious peak frequencies and the phenomenon of spectral leakage. Key–Words: Burg algorithm, Spectrum estimation, Prediction error energy, Levinson-Durbin recursion.