摘要:
提出一种新的独立成分分析算法,在利用直方图估计概率密度函数的基础上,由极大似然函数法构造独立信号的特征,并且在估计概率密度函数时,对相应的阶梯函数采用磨光处理,引入参数 ,并证明了 的选择依赖于信号的统计特征以及采样的样本总数。模拟实验结果表明,该算法能提高信号干扰比。
关键词:
独立成分分析,
直方图估计,
磨光函数,
极大似然估计,
Parzen窗
Abstract:
A new Independent Component Analysis(ICA) algorithm is introduced by histogram estimation and maximum likelihood estimation. When estimating the probability density function, the corresponding step function is smoothed using parameter . It is proved that depends on statistical characteristics of signals and total number of samples. Simulation results demonstrate that signal-to-interference ratio has significant improvement.
Key words:
Independent Component Analysis(ICA),
histogram estimation,
smoothing function,
maximum likelihood estimation,
Parzen window
中图分类号:
龚丹丹, 刘国庆. 基于极大似然Parzen窗的独立成分分析[J]. 计算机工程, 2010, 36(18): 279-281.
GONG Dan-Dan, LIU Guo-Qiang. Independent Component Analysis Based on Maximum Likelihood Parzen Window[J]. Computer Engineering, 2010, 36(18): 279-281.