An improved empirical mode decomposition based on the combination of extreme learning machine and mirror extension for restraining the end effects

In the progress of empirical mode decomposition, there is an open problem called end effects. To solve this problem, an extrema extension method based on the combination of extreme learning machine and mirror extension is proposed in this paper. The extrema extension work includes two steps: firstly, the extreme learning machine method is utilized to predict several extreme points separately at both ends of the original data series to form the preliminary expansion signal; then the preliminary signal is further expanded by the method of extrema mirror expansion. From the final resulting signal the relatively true envelopes of the signal can be obtained and the end effects will be effectively resolved. The proposed method is applied in the processing of simulation and the cavitation signals. Compared with the traditional methods, the result of the proposed method shows its effectiveness and superiority in restraining the end effects.