Extracting trend of time series based on improved empirical mode decomposition method

Solving overshoot and undershoot problems existed in the spline interpolation of empirical mode decomposition (EMD), improving this method and extracting trend of time series with it are the main tasks of this paper. A method is devised by using simple means of successive extrema instead from the envelope average to form the mean envelope. In this way, only one spline interpolation is required rather than two during the course of sifting process of EMD. It is easier to implement, those problems can be alleviated and EMD method is improved. How to get the successive extrema of series and how to realize trend extraction are also expounded in the paper. Experimental results show that the improved EMD method is better at trend extraction than the original one.