中国空间科学技术

• • 上一篇    下一篇

基于前向神经网络的非线性平滑滤波器设计研究

盛守照,王道波,黄向华   

  1. 南京航空航天大学,南京航空航天大学,南京航空航天大学 南京210016 ,南京210016 ,南京210016
  • 发布日期:2005-04-25

Research on Design of Nonlinear Smoothing Filter Based on Feedforward Neural Networks

Sheng Shouzhao Wang Daobo Huang Xianghua ( Nanjing University of Aeronautics and Astronautics, Nanjing 210016)   

  • Online:2005-04-25
  • Supported by:
    航空基金资助项目(01C52015)

摘要： 文章阐述了基于前向神经网络的非线性平滑滤波器的设计问题,分析了上述平滑滤波器的存在性及其滤波误差方差的组成,给出了有限观测序列下网络学习的性能指标,由此提出了一种次优网络学习算法,它具有很好的抑制白噪声能力。

关键词: 前向神经网络, 非线性, 平滑滤波器, 航天器

Abstract: The design of nonlinear smoothing filter based on feedforward neural networks is discussed. The existence of the nonlinear smoothing filter and the composition of filter error variance are analyzed theoretically. The target function for training networks is given, then a new suboptimal algorithm for training networks is proposed, which has good noise suppression capability. Its reliability and advantage are illustrated through an example.