基于模糊神经网络的自适应滤波方法仿真研究

提出了一种模糊自适应卡尔曼滤波算法。该算法基于模糊规则，根据新息相关性，自适应调整测量噪声方差R，有效的解决了噪声的统计特性与实际不符时，滤波器发散的现象。同时，利用Elman网络作为误差估计器，补偿模糊自适应卡尔曼滤波器的估计误差。仿真结果表明，两种方法结合，可以有效地防止滤波器发散，缩小实际的滤波误差，提高滤波精度，实现滤波器参数的在线改进。

Research on Adaptive Filtering Based on Fuzzy Neural Networks

An algorithm of adaptive fuzzy Kalman filtering is presented.According to relativity of innovation,the measurement noise covariance(R) was adjusted adaptively based on the fuzzy rule.The method couldcope with divergence problem caused by the insufficiently knowing of the prior filter statistics.Simultaneity,the Elman network was employed as a compensating error estimator to compensate estimation error of the fuzzy Kalman filter.Simulation result shows that the presented method reduces the error of actual filter,improves the accuracy and can amend filter's parameters on-line.