An Adaptive Analysis Method for Non-stationary Data_Empirical Data Decomposition

An adaptive analysis method for non-stationary data called empirical data decomposition (EDD) is proposed in this paper. The structures and design methods for analysis filters and synthesis filters are presented. Finally, a design method for adaptive prediction filter is discussed. The analysis filter is determined automatically by the observation data, and it is not a linear filter in which the parameters are not constant. The analysis algorithm can also implement multi-resolution analysis just like the wavelet transform. EDD is suitable for the decomposition of non-stationary data, and can wipe off the correlations between the observation data efficiently, so valid to describe the data with shorter symbol. Simulations are performed and the experimental results show the entropy of the coefficients produced by EDD is less than that by CDF5/3, so it is an effective analysis method for non-stationary data.