Unmanned Noise Reduction Method of Micro-Electro-Mechanical System Inertial Measurement Unit Based on Improved EMD

The traditional Fourier algorithm cannot fix the problem of non-stationary noise deduction for MEMS-IMU, therefore this article uses Empirical Mode Decomposition (EMD) algorithm to denoise the signal. In this article, Extreme Learning Machine (ELM) is combined to reduce the influence of end effect in the decomposition. First, the MEMS-IMU simulate signal generated by matlab is taken as the test object, and EMD as well as ELM extension decomposition are carried out for it respectively. The decomposition noise reduction effect is compared and analyzed to study the role of EMD and ELM in the process. Second, carry out a test on the authentically measured MEMS-IMU signal. In de-noising the MEMS-IMU simulate signal and authentic signal, we can analyze the noise deduction effect and observe the changes of parameters related to random error. The results show that the method based on ELM and EMD can achieve good noise reduction effect for MEMS signal.