High resolution wavelet estimation by ARMA modeling
Zhang Ya-nan1, Dai Yong-shou1, Wang Shao-shui2, Peng Xing1, Niu Hui1
1. College of Information and Control Engineering, China University of Petroleum (East China), Dongying, Shandong 257061, China;
2. The 41st Institute of China Electronics Tech-nology Uroup Corporation, Qingdao, Shandong 266555, China
Abstract:The most importantly advantage of ARMA(autoregressive moving average) model is to describe an exact wavelet with fewer parameters.Order over-determination easily leads to large calculation costs while order under-determination cannot meet the wavelet requirements.Higher-order cumulants are sensitive to special slices and it causes poor results with short time data series.This paper focuses on the model order determination.Singular value decomposition(SVD) based on autocorrelation function is exploded to determine the AR model order.Combining the information theoretic criteria method with the cumulant-based method,the author proposes a new MA model order determination method.Numerical simulations and real data processing show that additional Gaussian colored noise is suppressed,and the MA model order determination precision is improved by the information theoretic criteria method.With this approach,the wavelet gets high resolution and the model order number is reduced as much as possible in order to improve the computation efficiency.