BLKF方法抑制MEMS惯性传感器随机噪声

针对微机电系统（MEMS）惯性传感器应用卡尔曼滤波（KF）处理数据时随机噪声难以估计的缺点，提出一种基于贝叶斯拉普拉斯卡尔曼滤波（BLKF）的随机噪声更新方法。该方法利用拉普拉斯近似算法，将贝叶斯边缘分布近似表示，能够避免贝叶斯方法更新过程中由于数据维数较大造成后验边缘分布较难估计的情况，提高对随机噪声的更新效率和精度。通过陀螺仪转台实验采集实验数据，运用KF与BLKF分别滤波。对比滤波结果可发现，BLKF比KF能更好地反映真实角速度状态。同时，在陀螺仪角速度变化时，BLKF的可靠性和准确性更高，滤波效果更具优势，能够达到抑制MEMS惯性传感器随机噪声的目的。

Suppression of Random Noise of an MEMS Inertial Sensor by BLKF Method

Aiming at the shortcoming that the random noise is hard to be estimated when applying the Kalman Filter (KF) to an MEMS inertial sensor, a random noise updating method based on the Bayesian Laplacian Kalman Filter (BLKF) is proposed. The Laplacian approximation algorithm is used to approximate the Bayesian edge distribution. This method can avoid simulating difficulties of posterior edge distribution due to the large data dimension in the Bayesian updating process, and improve the update efficiency and accuracy of the random noise. The experimental data are collected by a gyroscope turntable experiment and filtered by the BLKF and KF respectively. Compared with the results of the filtering, we can find that the BLKF can better reflect the real angular velocity state than the KF. At the same time, the BLKF has higher reliability and accuracy when the angular velocity of the gyroscope is changed, so the filtering effect is more advantageous and the random noise of the MEMS inertial sensor can be suppressed.