计算机工程与应用 ›› 2010, Vol. 46 ›› Issue (27): 128-130.DOI: 10.3778/j.issn.1002-8331.2010.27.035

• 数据库、信号与信息处理 • 上一篇    下一篇

改进的快速独立分量分析方法研究

赵立权1,贾雁飞2   

  1. 1.东北电力大学 信息工程学院,吉林省 吉林市 132012
    2.北华大学 电气与信息工程学院,吉林省 吉林市 132013
  • 收稿日期:2010-05-24 修回日期:2010-07-08 出版日期:2010-09-21 发布日期:2010-09-21
  • 通讯作者: 赵立权

Research of improved fast independent component analysis

ZHAO Li-quan1,JIA Yan-fei2   

  1. 1.College of Information Engineering,Northeast Dianli University,Jilin 132012,China
    2.College of Electric and Information Engineering,BeiHua University,Jilin 132013,China
  • Received:2010-05-24 Revised:2010-07-08 Online:2010-09-21 Published:2010-09-21
  • Contact: ZHAO Li-quan

摘要: 快速独立分量分析算法硬件实现困难,基于Huber M估计函数的独立分量算法硬件实现容易,但是该算法稳健性有待提高。针对这一问题,提出了一种硬件实现容易,而且具有很好稳健性的改进的快速独立分量分析算法。该算法硬件实现容易,且具有很好稳健性的Tukey双权函数作为原算法代价函数中的非线性函数,通过牛顿迭代方法对代价函数进行优化,得到改进的快速独立分量分析算法。仿真实验证明,该算法是有效的,且稳健性较好。

Abstract: Fast independent component analysis algorithm is difficult to be realized by hardware,while independent component analysis based on Huber M estimator is easy to be realized by hardware,but its robustness need to be improved.To overcome the problem,a new fast independent component analysis algorithm which can be easily realized by hardware and is more robust is proposed.This new algorithm uses Tukey function which is easy to be realized by hardware and is robust as nonlinear function of cost function,and deduces the improved algorithm by using Newton iterative method to optimize the cost function.Simulations prove the correctness and robustness of the algorithm.

中图分类号: