Signal feature extraction based on cascaded multi-stable stochastic resonance denoising and EMD method

On the basis of cascaded multi-stable stochastic resonance system (CMSRS) theoretical studies, for the empirical mode decomposition (EMD) in heavy noisy mixtures, a method of EMD based on CMSRS denoising is presented. First, CMSRS is employed as the pretreatment to remove noise by virtue of its good effect in denoising performance, and the energy gradually is shifted from high to low frequency, then the denoised signal is decomposed by EMD. In simulated experiment, EMD is used to decompose the original and CMSRS output signals respectively. The result from the comparison shows that this method, not only removes high-frequency noise efficiently, but also reduces the decomposition layers and lets them have more reality meanings. At last, a diagnosis on the fault of inner race of rolling bearing confirms that this method removes high-frequency noise step by step, improves low-frequency signal's energy, and can effectively identify characteristic signals.