基于经验模态分解的光纤陀螺随机序列平稳化处理

FOG Random Sequence Pretreatment Using Empirical Mode Decomposition

采用AR模型等时间序列分析方法对光纤陀螺随机误差进行建模前需要对传感器输出序列进行平稳化检验和处理。引入经验模态分解法,对低精度光纤陀螺输出序列进行平稳化处理,采用逆序法检验平稳化处理的效果,引入偏态和峰态系数对随机系列进行正态性检验,并与小波分析等方法进行比较实验。结果表明:经验模态分解法不仅能有效去除陀螺输出信号的趋势项,还能提高信号的正态性。处理后的信号可以采用AR模型进行建模。

The FOG output sequence must be tested and treated before modeling with AR model.The empirical mode decomposition（EMD）method is introduced for removing the trend term of nonstationary random sequence.An inversed order method is used to check the effect of EMD method,and the skew value and kurtosis value are also introduced for normality test.The experiment indicates that the EMD method can not only eliminate the trend term of nonstationary random sequence,but also improve the normal natures of the nonstationary random sequence.The treated random sequence can be formulated by AR model.