一种适用于微弱信号提取的盲源分离算法  

A blind separation algorithm for weak signal extraction

作　　者：理华[1,2] 侯朝焕[1] 马晓川[1] 杨俊[1,2] 

机构地区：[1]中国科学院声学研究所,北京100190 [2]中国科学院研究生院,北京100039

出　　处：《应用声学》2009年第4期249-253,共5页Journal of Applied Acoustics

摘　　要：本文首次提出了适用于微弱信号提取的盲源分离算法,这种方法是在常用的自然梯度串行更新算法基础上采用了有监督的机制。通过观察两个信号互相关的程度自动判别激活函数的种类,实现了超高斯亚高斯强弱信号混合下的盲源分离。信号仿真表明,有监督的盲源分离技术收敛速度较快,精度较高。A blind separation algorithm for weak signal extraction is proposed. This algorithm is based on the natural gradient algorithm but in addition supervises the quality of separation to estimate the activation function automatically on noting the correlation of two signals. Mixed super-Gaussian and sub-Gaussian signals of which one is strong and the other is weak are separated successfully. The simulation shows that this algorithm converges relatively precisely and quickly.

关 键 词：盲源分离 独立分量分析 超高斯 亚高斯 激活函数 

分 类 号：TN912.3[电子电信—通信与信息系统] TP391.4[电子电信—信息与通信工程]