To reduce the random drift in Fiber Optic Gyro (FOG),a self-adaptive filtering method called variable step-size sign Least Mean Square (LMS) algorithm was presented. Considering the practical engineering requirement,some positive work was carried on based on the traditional LMS algorithm. On one hand,normalized power value of input signal was introduced to improve filtering accuracy and stability; on the other hand,sign function of error was used as well in order to decrease calculation burden and increase real-time performance. Besides,the step-size parameter was determined by convergence analysis. The actual static and rotation experiment results indicate that the new algorithm is able to select the step-size parameter in advance and with more favorable filtering stability,accuracy and real-time performance. The variable step-size sign LMS filtering method is suitable for engineering application.