Abstract: The empirical mode decomposition (EMD) method presents its own ability for processing nonlinear and non-stationary signals. It can effectively obtain the time-frequency characteristics of non-stationary signals. But there is an involved end effect in the course of getting two envelops of the data using spline interpolation. Based on the consideration of the mechanism of the end effect, a new method for restraining the end effects of B-spline empirical mode decomposition (BS-EMD) based on the neural network ensemble is proposed. The data extension technology based on the neural network ensemble is described. The two ends of the original signal are extended and predicted using the neural network ensemble method. Then, the mean interpolation curve of the extended signal is calculated by B-spline interpolation function. The intrinsic mode functions are calculated by EMD. The results of simulation and practical signal analysis show that the method can restrain the end effects of BS-EMD.

Key words: 国家自然科学基金(51105323)和河北省自然科学基金(E2012203166;F2009000500)资助项目