A wavelet method is introduced to compensate stochastic errors of FOG attitude measurement system, analysing and comparing threshold and threshold function principle, a new threshold function combining with soft and hard threshold was designed. Adopting data moving window realizes recursive computation and real-time data update and solves filtering signal’s boundary problem with symmetric periodic extension method, and achieves real-time wavelet threshold filter. Experimental results show the feasibility of the real time wavelet threshold method in FOG signal de-noising, the real-time wavelet de-noising algorithm with new threshold function has the advantages of single algorithm. which improves the precision and stability of the FOG.