作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程

• 移动互联与通信技术 • 上一篇    下一篇

有界混合信号的快速分离算法

苏巧,沈越泓,徐鹏程   

  1. (解放军理工大学通信工程学院,南京 210007)
  • 收稿日期:2015-01-26 出版日期:2016-02-15 发布日期:2016-01-29
  • 作者简介:苏巧(1990-),男,硕士研究生,主研方向为移动通信;沈越泓,教授、博士生导师;徐鹏程,博士研究生。
  • 基金资助:
    国家自然科学基金资助项目(61172061);江苏省自然科学基金资助项目(BK2011117)。

Fast Separation Algorithm for Bounded Mixed Signals

SU Qiao,SHEN Yuehong,XU Pengcheng   

  1. (College of Communication Engineering,PLA University of Science and Technology,Nanjing 210007,China)
  • Received:2015-01-26 Online:2016-02-15 Published:2016-01-29

摘要: 为从有界混合信号中分离出源信号,提出一种检测和移除稳态波动的有界成分分析算法。该算法基于信号集合紧性和笛卡尔可分性的假设,在不考虑源信号估计的幅度、排列和相位不确定性的条件下,完成独立源和非独立源的分离,对学习曲线的稳态波动首先利用曲线稳态处的相关性进行检测,再使用变步长迭代的方法移除学习曲线的稳态波动,并给出算法停止准则,从而提高算法收敛速度和分离精度。仿真结果表明,该算法能有效分离相关源和非相关源信号,且无论是在无噪声或有噪声的条件下,相对于现有的有界成分分析算法收敛速度更快,精度更高。

关键词: 盲源分离, 有界成分分析, 稳态波动, 超椭球, 超矩形

Abstract: In view of the separation for the mixture of bounded signals,this paper proposes a Bounded Component Analysis(BCA) algorithm for detection and removing of steady state fluctuation.The algorithm takes advantage of assumptions of compactness and Cartesian decomposition of convex support of sources.It can separate both independent and dependent sources up to permutation,scaling and phase ambiguities.The algorithm detects the fluctuation of the steady state learning curve by the correlation at the steady state,uses the iteration of variable step to remove the fluctuation of the steady state learning curve.This leads to a faster convergence speed and better performance.Simulation results show that the proposed algorithm can separate the dependent and independent sources efficiently and the convergence speed and accuracy are superior to existing BCA algorithms in the noisy or noiseless circumstance.

Key words: Blind Source Separation(BSS), Bounded Component Analysis(BCA), steady state fluctuation, hyper-ellipsoid, hyper-rectangle

中图分类号: