The data collected from the metro stray current monitoring system based on optical fiber sensing existed noise，which mainly included three aspects，such as the noise of sensing optical path system，the noise of photoelectric conversion and the noise of signal acquisition circuit. In order to eliminate efficiently those noises from the whole sensing system，the wavelet analysis is used to process signal-noise separation. The wavelet threshold denoising method is used，the threshold function is reasonably chosen and Daubechies wavelet is selected as a wavelet basis，which is used to eliminate the noise of current signal and stray current signal. The results show that the wavelet threshold denoising method can eliminate most of the noise，the curve is ideal after processing，and the accuracy of the measurement is ensured.