Noise is one key factor to affect the precision of fiber Bragg grating (FBG) sensor system for demodulating the centra l wavelength.The wavelet transform can get good de-noising performance due to i ts good characteristics of time-frequency and the multi-resolution.By analyzin g the features of typical soft and hard threshold methods applied for wavelet de -noising,the drawback is to affect the system measurement precision.Moreover,th e soft and hard threshold methods need to optimize the threshold function′s adju stable factor which is difficult to be determined.An improved threshold function without adjustable factor is put forward for wavelet transform in this paper.In tegrating the function with the translational invariant (TI) wavelet de-noising method,an application of de-noising the FBG sensor signal is studied to overc ome the pseudo-Gibbs phenomena.The simulation and experimental results show tha t the translational invariant wavelet with improved threshold function can deal with FBG sensor signal with different signal-to-noise ratios (SN R from 5dB to 40dB),and can get the best SNR and root of mean square error (RMSE) simul taneously.Com pared with the traditional soft threshold method,its average SNR increases by 3dB,while compared with the reference improved thresh old method,its average SNR improves by 2dB.So,the sy stem can meet the requirement of improving FBG sensor detection precision.