A de-noising method for vibration signals based on compressed sensing

A de-noising method for vibration signals based on compressed sensing theory is proposed. First the original signal is projected into a low-dimensional space. Considering the fact that the uncontaminated vibration signal can be represented sparsely by some dictionary matrix while noise signals can´t be, then the noise information in original signal would be discarded in compressed projection. Then the original signal can be reconstructed with orthogonal matching pursuit algorithm and de-noising is achieved finally. The method is validated with gear simulation signal contaminated by Gaussian noise. The de-noising results are analyzed with different sparsity set in orthogonal matching pursuit algorithm and different measurements set in projection. The test results show that the SNR could be improved obviously when setting suitable sparsity and measurements, which inferring that the proposed method should be effective in vibration signal de-noising.