针对标准Kalman滤波需要建立准确的系统模型、小波阈值降噪对小波基和阈值的选取依赖于经验的不足，将局域均值分解(LMD)方法引入MEMS陀螺的随机误差滤波。该方法可自适应地将随机误差信号分解为若干PF分量之和，且对各分量进行小波降噪处理，将处理后的各分量相加得到降噪信号。实验分析表明该滤波方法效果明显。

The standard Kalman filter needs exact system model,and the choosing of wavelet base and threshold in wavelet threshold denoising is dependent on the experience.Taking the problems into consideration,we proposed a new filtering method based on Local Mean Decomposition (LMD).Random error signal was decomposed to several Production Function(PF) adaptively,and wavelet denosing was made to some given frequency PF that contained noises.Each PF after processing can be reconstructed to obtain the denoised signal.The experiment shows that wavelet threshold denoising based on LMD has obvious effect.