基于小波包变换与前向线性预测滤波的光纤陀螺去噪算法

Denoising algorithm for FOG based on wavelet packet transform and FLP algorithm

为了减小环境噪声对光纤陀螺输出的影响,提出了一种新的基于小波包变换(WPT)和前向线性预测滤波算法(FLP)的去噪方法———WPT-FLP算法.首先介绍了小波包分解与FLP算法的原理,并对2种算法进行了融合,利用小波包变换进行信号的分解与重构,在此过程中对小波包分解后的高频系数进行强制去噪,对小波包分解后的低频系数进行FLP去噪,从而在有效保留信号中有用成分的同时,显著地提高了信号的去噪精度.最后利用静态下的光纤陀螺信号对WPT-FLP算法进行了验证.验证结果及Allan方差分析表明:与传统算法相比,WPT-FLP算法的去噪精度更高,原始信号的标准差由5.1×10-3(°)/s减小到0.22×10-3(°)/s,Allan方差分析噪声误差项中的量化噪声与角度随机游走也分别由0.47×10-6 rad和4.2×10-3(°)/h1/2减小到8.1×10-8 rad和0.14×10-3(°)/h1/2.

In order to reduce the impact of environmental noise on the output of fiber optic gyroscope（FOG）,a new denoising algorithm（named WPT-FLP algorithm） based on forward linear prediction algorithm（FLP） and wavelet packet transform（WPT） is proposed.The denoising principle of FLP and WPT is introduced first,then the advantages of the two algorithms are fused.WPT is employed to decompose and reconstruct the signal,and conduct compulsory denoising at high frequence coefficients;meanwhile,the FLP algorithm is employed to conduct denoising at low frequence coefficients.Thus,the denoising accuracy can be improved greatly while the useful ingredients of the signal can be reserved effectively.Finally,the static signal of FOG is employed to verify the proposed algorithm.The simulation results and Allan variance analysis show that compared with traditional denoising methods,the standard deviation of original signal is decreased from 5.1×10-3 to 0.22×10-3（°）/s,the quantization noise and angle random walk of Allan variance analysis are decreased,respectively,from 0.47×10-6rad and 4.2×10-3（°）/h1/2 to 8.1×10-8rad and 0.14×10-3（°）/h1/2 by using the proposed denoising algorithm.