Laser Gyro Signal De-noising Based on EMD and Kernel Principal Component Analysis

In order to suppress the random shift error of laser gyro and improve the practical precision of inertial navigation systems，an improved gyro denoising method is proposed that combines empirical mode decomposition(EMD) and kernel principal component analysis(KPC'A).In the proposed algorithm，a gyro signal is decomposed as a series intrinsic mode function（IMFs) by EMD. In turn，the noise energy contained in each IMF is approximately estimated by using the IMF noise energy distribution model，and then，decomposing the each IMF by KPCA，and adaptively selecting the principle components which are should be retained. At last，the denoised gyro signal is obtained by accumulating the each processed IMF by KPCA. A detailed comparison between the proposed method and the wavelet methods is given. The denoising effect of different methods is analyzed by the overlapping Allan variance. Experimental results show that the proposed method performs better in removing noise than classic wavelet methods and can more efficiently suppress the gyro random drift.