阈值去噪与RBF神经网络在MEMS陀螺仪误差补偿中的应用

Application of Threshold Denoising and RBF Neural Network in the Error Compensating of MEMS Gyro

针对现有MEMS陀螺仪中随机误差较大,导致器件输出信噪比低进而影响其应用范围的现状,提出一种基于小波阈值去噪与梯度径向基(RBF)神经网络结合的MEMS陀螺漂移非平稳时间序列建模预测方法。首先采用Allan方差法分析了MEMS陀螺仪的主要随机误差,随后利用小波阈值去噪分离出MEMS陀螺误差模型中的白噪声及漂移误差,最后采用RBF神经网络对漂移数据进行建模。通过实验对文中所述的误差补偿方法进行验证,表明了方法的有效性,对于基于MEMS陀螺仪的惯导系统精度提高具有重要意义。

The large random errors in MEMS gyros have resulted in low signal noise ratio in devices outputs,which affects the application scope of gyros in turn. To solve this problem,this paper proposed a modeling forecasting method towards MEMS gyro drift using non-stationary time series, this method is based on wavelet threshold de-noising combined gradient radial basis（RBF）neural network. Principal random errors of MEMS gyros is analyzed using Allan variance,then the wavelet threshold de-noise is used to separate white noise and drift error in error model of MEMS gyros,drift data finally modeled with RBF neural network. Experiments were carried to validate the error compensate method proposed in this paper, the experimental results show the effectiveness of the method, there is great significance for improving the precision of inertial navigation system based on MEMS gyroscope.