As the laser gyroscopes and accelerometers in a ring laser gyroscope Strapdown Inertial Navigation System(SINS) generate high noises by the dither motion of a base and a low pass fitering and a wavelet method can not suppress the noises in real time effectively,this paper presents a prefiltering method combining a low pass filtering and a Kalman filtering based on a hidden Markov model.Firstly, the output of a sensor is filtered by the lowpass filter,then it is filtered by the steady-state Kalman filter based on the hidden Markov model.By this way, the large sensor noise brought by the dither motion of the base can be lowed down to a very low level. Experiment results show that this prefiltering method can work efficiently with a low computational complexity. When the vehicle engine is on, the standard variances of the ring laser gyroscope and accelerometer are suppressed from 300°/h to 1°/h and from 11 mg to 40 μg, respectively. It is concluded that the proposed prefiltering can help SINS accomplish the initial alignment on a rocking base.