Abstract: The traditional empirical mode decomposition (EMD) usually uses cubic spline interpolation to fit the upper and lower envelopes of the signal, leading to ill behavior on edges and therefore degrading the signal processing quality. We resolve the edge problem of EMD by using the second-generation wavelet denoise technique. Compared with the first generation wavelets, the second generation has lower computational complexity and less memory requirement. Numerical experiments show that the second-generation wavelet denoising can effectively reduce the edge effect with fewer decomposition layers, making EMD more practical.

Key words: empirical mode decomposition (EMD), cubic spline, edge effect, second generation wavelet