摘要
在信号滤波算法优化问题的研究时,扩展卡尔曼滤波算法的精度依赖于系统模型精确性。采用一种改进的扩展卡尔曼滤波算法研究了状态量和观测量相同的系统。用滤波后的状态量和当前观测量以得到实时噪声,求出过程噪声方差阵,在此基础上利用支持向量回归机算法对过程噪声方差阵进行训练,从而得到新的过程噪声方差阵,用此阵作为下一次扩展卡尔曼滤波的过程方差阵,以后继续上述迭代过程。结果证明方法极大的提高了滤波精度。仿真说明方法的有效性。
Concerning the defect that the accuracy of the extended Kalman filter（EKF） depends on the accuracy of system model,in this paper,we use an improved EKF algorithm to research a system whose state variables are equal to observing variables.Firstly,we use the filtered state variables and current observing variables so as to obtain the real time noises,from which we then find the process noise variance matrix.On this basis,we train the process noise variances by using support vector regression（SVR） to acquire new ones,which we consider as the process variance matrix in the EKF next time.Next,we continue this iterative process.With this method,the accuracy of filter is improved greatly.At the end of this paper,we present some concrete simulative examples to show the validity of this method.
出处
《计算机仿真》
CSCD
北大核心
2011年第4期156-159,175,共5页
Computer Simulation
基金
国家自然科学基金项目(60574004)
国家自然科学基金资助重点项目(60736024)
教育部科技创新工程重大项目培育资金项目(708069)
关键词
扩展卡尔曼
噪声方差阵
支持向量回归机
滤波精度
Extended Kalman filter（EKF）
Process noise variance matrix
Support vector regression（SVR）
Accuracy of filter
