A Parallel Denoising Model for Dual-Mass MEMS Gyroscope Based on PE-ITD and SA-ELM

In a bid to solve the gyroscope temperature drift problem, a parallel denoising model based on PE-ITD and SA-ELM has been proposed in this paper, wherein, Intrinsic time-scale decomposition (ITD) is an effective signal decomposition algorithm, Permutation entropy (PE) is a entropy to accurately determine the complexity of the signal, Extreme Learning machine (ELM) is a machine learning algorithm for predicting, and Simulated Annealing(SA) for finding the optimal parameter set. First, ITD is employed to decompose the output signal of the gyroscope and the PE can distinguish the decomposition results into noise-only component, mixed component and trend component. Secondly, the mixed component is filtered by Forward liner prediction (FLP), the denoised signal can be obtained after processing. Finally, the trend component is compensated by SA-ELM, through reconstruction, the compensation result can be obtained. As is shown in experimental results, the parallel model proposed in this paper is able to effectively eliminate the temperature error compared with the traditional serial model. Compared with the EMD analysis method, the ITD is superior to the EMD analysis method in terms of computational efficiency and instantaneous information. The proposed SA-ELM algorithm takes simulated annealing algorithm to find hidden layer neurons optimal number adaptively, which can improve the models accuracy and generalization. This paper demonstrates the superiority of this method.