Abstract:  In order to realize the intellectualization of mining unit，the position and posture of mining machinery must be controlled in real time and accurately. Because the mining working face environment is complex， strapdown inertial navigation technology can be used to monitor mining machinery.As the core component of strapdown inertial navigation system， the random noise of fiber optic gyro restricts the precision of inertial navigation system.Traditional modeling filtering and wavelet denoising methods can not guarantee the accuracy and denoising result because of their own limitations. Combined with empirical mode decomposition （EMD），a denoising algorithm combining Hausdorff distance （HD） screening criteria and threshold filtering is proposed. In this algorithm， HD of probability density functions （PDF） which are the estimate of intrinsic mode function （IMF） and the original signal is used as the discrimination basis to screen all IMFs，and then thresholding is introduced to filter the selected IMFs and reconstruct it with remaining terms.The filtering results of soft and hard thresholds is compared through tests and the hard threshold are adopted in the algorithm.In order to verify the effectiveness of the algorithm， the algorithm is compared with EMD-Cor，EMD-CMSE and other methods.Simulation and test results show the superiority of the method， which can effectively reduce the random errors of gyro.

Key words: FOG, Random drift, EMD, Hausdorff distance, Threshold filtering