【作者】 张秀艳；

【导师】 王秀芳；

【作者基本信息】 大庆石油学院 ， 电力电子与电力传动， 2003， 硕士

【摘要】 本论文对神经网络理论应用于最优非线性滤波进行了研究，主要完成了以下的研究工作： 分析了神经网络理论应用于最优非线性滤波的现状及发展趋势，并对神经网络和经典的最优非线性滤波方法进行简单的讨论，其中包括非线性最小方差(LMS)估计(即扩展的卡尔曼滤波)和非线性最小二乘估计(LS)，研究了神经网络应用于最优非线性滤波的可行性。 探讨了基于反向传播(DP)网络和基于径向基函数(RBF)网络的最优非线性滤波，利用MATLAB作为仿真软件，并应用BP网络和RBF网络对实例进行了仿真，得出一系列仿真波形，从实验的角度验证了该模型仿真设计思想的正确性。 最后部分通过计算机仿真实验，验证了RBF网络在非线性滤波方面的优越性，它具有训练时间短、所用神经元数目少、精度高等突出优点。可以看到，神经网络用于最优非线性滤波具有广阔的发展前景。更多还原

【Abstract】 This paper explores the neural networks approach to optimum nonlinear filtering, the main research work is as follows:It analysis the actualities and the developing trend about the neural networks approach to optimum nonlinear filtering, and it surveys neural networks and the classical theory about the optimum nonlinear filtering. Among them include not only the nonlinear Least-Square (LMS) estimation (spread of Kalman filter) but also the nonlinear Least Mean Square estimation (LS). At the same time it analysis feasibility of integration of neural networks and optimum nonlinear filtering.It discusses the Back Propagation (BP) network and the Radial Basis Function (RBF) network to optimum nonlinear filtering, and it makes use of MATLAB as simulation software, and applying the BP network and the Radial Basis Function (RBF) network simulate realism examples, it obtain a series of simulation waves, thus , the accuracy of the simulation model is validated from the manner of experiments.Finally, it by the computer simulation verifies the superiority for the RBF’s in nonlinear filtering, the RBF have many advantage such as shorter training time, least note and higher accuracy. It can be seen, the ANN used in the optimum nonlinear filtering will be the tendency of the filtering fields.更多还原