To demodulate the central wavelength precisely is the key of FBG sensor sy stem.How to de-noising is a challenging problem.In order to gain the drift of wavelength and achieve the me asurement of ambient environment parameters precisely,nonlinear de-noising methods are investigated in this paper.Wavelet threshold de-noising is usually a method of nonlinear de-noising.The selection of threshold function and threshold value is important in wavelet threshold de-noising.Firstly,this paper analyze s the problem of choosing the threshold function and threshold in traditional soft and hard threshold de -noising,and then improves the self-optimizing threshold.Secondly,the paper proposes a method of gaining the o pt imal value of a (regulatory factor) in the compromising threshold function.Simulation and experiment results show t hat the signal-to-noise ratio (SNR) changes with a.When the value of a changes betwee n 0.40and 0.45,the effect of de-noising will become better. Compared with the traditional soft threshold and hard threshold methods,the SNR is increased by 2－4dB,and the mean square root error (MSE) is reduced by 0.2－0. 5.While compared with the reference improved threshold method,the SNR is improved by 1dB and the M SE is reduce d by 0.1.The overall de-noising effect is better than t he others.