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.
出处
《应用声学》
CSCD
北大核心
2009年第4期249-253,共5页
Journal of Applied Acoustics
关键词
盲源分离
独立分量分析
超高斯
亚高斯
激活函数
Blind separation, Independent component analysis, Super-gaussian, Subgaussian, Activation function.
