Abstract: A noise reduction method was proposed based on the chaotic synchronization system, and was applied to process the vibration signal of the rolling bearing and to diagnose the bearing fault in combination with the power spectral density. Firstly, analyzing the noise reduction mechanism of the chaotic synchronization system, the influence of the input signal on the phase space trajectory was discussed for different running state. Then, a noise reduction model of the chaotic synchronization system was built based on the Chua's circuit, analyzing its characteristics and comparing with other methods. Finally, taking the actual vibration signals of three different damage modes of rolling bearing as the input of the model, the waveform and the change of power spectrum density of vibration signal and synchronous error signal were compared based on their respective fault characteristic frequencies. Taking advantages of the noise immunity and synchronism of the chaotic system, the problems of parameter setting complexity and the decision difficulty of the system running state can be avoided in the existing chaotic detection methods.The experimental results show that the new chaotic synchronization system can effectively improve the signal to noise ratio of the measured signal, and is suitable for the pre-processing of the signal. And it can be combined with the traditional method to form a new fault diagnosis method.