Electrostatic sensor was implemented in roller bearing experiment investigation to monitor defects. In consideration of various strong noises involved in the electrostatic monitoring of roller bearings and the difficulty to obtain defect characteristics

a united denoising method was put forward based on spectrum interpolation and difference spectrum of singular value. Firstly

the spectrum interpolation was used to eliminate the power line interference; then the resultant signal was used to construct Hankel matrix; the number of useful components was automatically selected based on the difference spectrum of singular value

and finally the signal was reconstructed. Simulation and practical experiments show that the useful components of the signals involving high power line interference cannot be extracted by performing difference spectrum of singular value method or the wavelet denoising method; the presented denoising method can highlight the defect characteristic frequency in early stage more effectively than the spectrum interpolation and wavelet denosing method.