ANALYSIS OF SIGNAL PREDICTION BASED ON EMD AND ANN
Wang Yong 1) ; Yang Jing 2) ;Zhang Lihui 2) ; and Zhang Hongjuan 3)
1)Earthquake Engineering Research Center of Hebei Province, Tangshan 063009; 2)College of Mining Engineering, Hebei United University, Tangshan 063009;3)Great Wall College, China University of Geosciences,Baoding 071000
Abstract:In view of that the EMD (Empirical Mode Decomposition, referred to as EMD) in dealing with nonlinear, nonstationary signal, and artificial neural network (Artificial Neural Networks, referred to as ANN, also known as neural network) can deal better with the nonlinear problems,we propose a new way combining these two methods for dealing with the signal prediction.Firstly, by using EMD to decompose the simulation signal with noise into several IMF(Intrinsic Mode Function, referred to as IMF) components and a tendency and then in two ways to deal with the endpoint problem in the decomposition process. The results show that the two methods can be good to solve the endpoint problem, then for each component using RBF (Radial Basis Function) neural network to predict separately, and reconstruct the final prediction results. Compared with the use of neural networks to predict directly without being processed with EMD and real data, this method has a higher prediction accuracy.