Single-channel bearing vibration signal blind source separation method based on morphological filter and optimal matching pursuit (MP) algorithm

Abstract

A novel method to solve the single-channel source separation problem is proposed, which is based on a morphological filter to remove noise and the optimal matching pursuit (MP) algorithm to create pseudo sources. The signal is first purified by the morphological filter. Then, the purified signal is decomposed by the MP method, the Gabor atoms of the MP are selected by genetic algorithm in order to be less time-consuming. The selected atoms and the purified signal are combined into multi-dimensional signals, and the underdetermined problem of single-channel source separation is solved. The singular value decomposition method is used to estimate the number of new constructed signals. Then, the fast independent component analysis algorithm is used to achieve the separation of the signals. The proposed method is applied to the separation of the simulated signals and the mixed faults of bearing, the results indicate that the method can well solve the single-channel signal separation problem.