Noisy blind source separation based on CEEMD and Savitzky-Golay filter

Abstract

The standard independent component analysis (ICA) algorithm is difficult to extract signals in noise condition, a blind separation algorithm based on denoising pretreatment was proposed. Mixed signals firstly were decomposed into several stationary intrinsic mode components (IMF) using complementary ensemble empirical mode decomposition (CEEMD), and high frequency IMF components were filtered with Savitzky-Golay filtering, then using the whole components reconstructed the mixed signals, finally applying the fast independent component analysis(FastICA) to separate the reconstructed signals. Simulation results showed that the proposed method improved the effect of blind signal separation under low signal-to-noise ratio.