EMD-EKF方法研究

Research on the EMD-EKF Algorithm

摘要:

应用扩展卡尔曼滤波(EKF)时需要估计量测噪声的统计特性。文中针对观测噪声统计特性描述不准确导致的EKF性能下降的问题,利用经验模态分解方法(Empirical Mode Decomposition,EMD)可以分离信号和噪声的特性,提出了一种在未知量测噪声条件下的EKF方法。该方法可以跟踪观测噪声的变化,即实现了对量测噪声的估计,从而解决了在未知量测噪声的情况下的EKF问题。仿真结果表明可运用于无源定位中。

Abstract:

It is well known that the successful application of EKF depends on whether the prior knowledge of the statistics of the measurement noise is known. In this paper, the EKF algorithm was first analyzed briefly. The feature of EMD of separating a noise signal into signal part and noise part was combined into EKF. A new method was then proposed that made the EKF under unknown measurement noise covariance condition valid. This presented method could track the changes of the measurement noise covariance in real time. Finally, the application of the proposed method for passive location was discussed. The simulation results verified the effectiveness of the proposed method.