Firstly, the signals derived by multi-micro-electro-mechanical systems(MEMS) gyroscopes are decomposed into
a series of intrinsic mode functions(IMF) based empirical mode decomposition method. At the same time, their instantaneous frequency and energy of IMF can be obtained through Hilbert transform. Then, a noise filter is constructed through analyzing nonlinear correlation coefficient between the IMF and the original signals, and applied to de-noise the signals of MEMS gyroscopes. The nonlinear correlation information entropy is used as the performance evaluation measure of signals fusion with different group of IMF’s, and the measurement accuracy of the fused signal can be increased. Finally, experimental results show the effectiveness of the proposed method.