De-noising stochastic noise in FOG based on second-generation DB4 wavelet and SURE-threshold

An effective de-noising method for fiber optic gyroscopes (FOGs) is proposed. This method is based on second-generation Daubechies D4 (DB4) wavelet transform (WT) and level-dependent threshold estimator called Stein’s unbiased risk estimator (SURE). The whole approach consists of three critical parts: wavelet decomposition module, parameters estimation module and SURE de-noising module. First, DB4 wavelet is selected as lifting base of the second-generation wavelet in the decomposition module. Second, in the parameters estimation module, maximum likelihood estimation (MLE) is used for stochastic noise parameters estimation. Third, combined with soft threshold de-noising technique, the SURE de-noising module is designed. For comparison, both the traditional universal threshold wavelet and the second-generation Harr wavelet method are also investigated. The experiment results show that the computation cost is 40% less than that of the traditional wavelet method. The standard deviation of de-noised FOG signal is 0.012 and the three noise terms such as angle random walk, bias instability and quantization noise are reduced to 0.007 2° /ℴh, 0.004 1° / h, and 0.008 1°, respectively.