Improved hybrid filter for fiber optic gyroscope signal denoising based on EMD and forward linear prediction

• We propose a novel EMD algorithm to solve mode mixing of noisy signal decomposition with less computational complexity.
• We verify that filtering method based on EMD can reserve more weak features of FOG drift output, and it is easy to implement.
• We verify that G-FLP can further reduce the residual noise contained in each stage of EMD.
• We extend the usage of FLP algorithm for complicated signal with small signal-to-noise ratio.

Fiber optic gyroscope (FOG) has been widely applied in strapdown inertial navigation system (SINS) as an ideal component. However the slowly varying drift of FOG that often submerged in noise will degrade the precision of SINS over time. The main objective of this paper is to focus on eliminating noise and extracting the slowly varying drift using a newly proposed hybrid filter called EMD-G-FLP. The implementation of EMD-G-FLP mainly consists of two steps. First, improved empirical mode decomposition (EMD) method is used to decompose original drift. Then a prediction filtering method named G-FLP is adopted to denoise obtained intrinsic modes. EMD-G-FLP is compared with methods based on wavelet packet translation (WPT) and G-FLP, respectively, using signals detected from a closed-loop interferometric FOG. The deficiencies of WPT-based method are analyzed by employing static and dynamic FOG drift. Experimental results show that, G-FLP and EMD-G-FLP retain the slowly varying drift without distorting the trend. Furthermore, compared with G-FLP, EMD-G-FLP reduces the noises including quantization noise, random walk and bias instability by about 82%, 75% and 53%, respectively.