计算机工程与应用 ›› 2013, Vol. 49 ›› Issue (4): 121-124.

• 网络、通信、安全 • 上一篇    下一篇

NGA实现互信息量最小化的盲源分离

陈  琛,马庆伦,李灯熬,赵菊敏   

  1. 太原理工大学 信息工程学院,太原 030024
  • 出版日期:2013-02-15 发布日期:2013-02-18

NGA achieves blind source separation with mutual information minimum

CHEN Chen, MA Qinglun, LI Deng’ao, ZHAO Jumin   

  1. Institute of Electronic Information Engineering, Taiyuan University of Technology, Taiyuan 030024, China
  • Online:2013-02-15 Published:2013-02-18

摘要: 提出了一种新的盲源分离算法,该算法通过自然梯度算法实现互信息量最小化,从而达到盲源分离的最佳效果。由于互信息量具有度量分离信号的循环相关矩阵和单位阵的相似程度的特性,最小互信量标志着分离矩阵最佳的状态。通过自然梯度寻优算法来实现互信息量的最小化,从而得到理想的分离矩阵。仿真结果表明算法对具有循环平稳特性的源信号分离效果显著,且收敛速度快。

关键词: 盲源分离, 循环平稳信号, 互信息量, 自然梯度算法

Abstract: A new proposed Blind Source Separation(BSS) algorithm with Natural Gradient Algorithm(NGA) achieves Minimum Mutual Information(MMI). It is reasonable to adopt mutual information to illustrate the similarity of the autocorrelation function and the information. To achieve the ideal separation matrix with MMI, the natural gradient algorithm is applied to the separation matrix optimization process. Simulation results verify the validity of the proposed algorithm. Simulation results show that the proposed algorithm has faster convergence speed and better separation effect than traditional gradient algorithm.

Key words: blind source separation, cyclostationary signal, mutual information, natural gradient algorithm