The Bi-dimensional Ensemble Empirical Mode Decomposition Analysis Based On Wavelet Function

Since N.E.Huang and his team proposed a HHT method which is about signal analysis,and HHT has attracted broad attention of scholars at home and abroad research.This method has achieved a series of research results because of profound study of scholars.Empirical Mode Decomposition is creative method which is the most important in HHT,and EMD is a kind of adaptive method of data processing or data mining,it is very suitable for nonlinear and non-stationary time series processing.So it is essentially a stationary processing of data sequence or signal.EMD received promptly and efficiently applications in different engineering fields.Adaptive is one of the most important features of the EMD algorithm,which is to smooth the nonlinear,non-stationary signal.EMD can decompose the complex data into finite Intrinsic Mode Functions(IMF)and a trend term,so that the concept of instantaneous frequency has a practical physical meaning.However,EMD has the problem of mode mixing.In order to overcome this shortcoming,an improved algorithm-the Ensemble Empirical Mode Decomposition(EEMD)method was proposed.It is a kind of noise assisted data analysis method proposed for the shortcomings of EMD method.In order to apply this algorithm to image processing,this paper proposes a two-dimensional empirical mode decomposition method based on spline wavelet function.The spline wavelet function is used as the basis function in the bi-dimensional algorithm.which implements adaptive data fitting using the least squares principle.In this paper,facial expression images as the object of study,proved that the proposed algorithm is feasible.This paper processes the facial expression images using EMD and bi-dimensional EEMD based on Wavelet Function,and we will draw a conclusion to compare those two methods.This paper chooses the 213 facial expressions from Japan JAFFE facial expression database to research and analysis.Firstly,we pre-process the facial expression images,which includes the method of scale normalized and histogram equalization.Secondly,this paper uses the method of EMD based on Radon transform and the method of bi-dimensional EEMD for face expression images further processing,respectively.Finally,we use SVM for the above results.That makes expression feature attribute data for training.This paper analyzes the results of the classification and compares to find more effective facial expressions that are identified.