基于概率密度核估计的ICA算法.DOC

题目：一种含参数的概率密度核估计的独立分量分析算法独立分量分析；互信息；评价函数；广义高斯分布；核估计分析了盲源分离算法中互信息准则与概率密度核函数的关系，利用广义高斯模型，提出了一种基于含参数的核概率密度估计的独立分量分析算法。该算法利用观测样本求峰度，通过分段函数给出相应高斯指数值，并利用样本数据进一步修正源信号的概率密度函数，实现对分离信号评价函数的精确估计。在此评价函数基础上，采用互信息最小化准则，推导出分离矩阵的迭代更新规则。所提算法在一定程度上解决了ICA算法中信号评价函数估计的难题，且能对任意源混合信号进行有效盲分离，仿真实验验证了算法的性能。刊号（CN、ISSN）一种含参数的概率密度核估计的独立分量分析算法 贠亚男1，郑 茂2， 郑林华2 （1.平顶山工业职业技术学院，平顶山 467001；2. 国防科学技术大学电子科学与工程学院，长沙 410073） 摘 要：分析了盲源分离算法中互信息准则与概率密度核函数的关系，利用广义高斯模型，提出了一种基于含参数的核概率密度估计的独立分量分析算法。该算法利用观测样本求峰度，通过分段函数给出相应高斯指数值，并利用样本数据进一步修正源信号的概率密度函数，实现对分离信号评价函数的精确估计。在此评价函数基础上，采用互信息最小化准则，推导出分离矩阵的迭代更新规则。所提算法在一定程度上解决了ICA算法中信号评价函数估计的难题，且能对任意源混合信号进行有效盲分离，仿真实验验证了算法的性能。 关键词：独立分量分析；互信息；评价函数；广义高斯分布；核估计 中图分类号：TP911.7 文献标识码：A 文章编号：1004-731X (2011) 11- Independent Component Analysis Algorithm Based on Parametric Kernel Probability Density Estimation YUN Ya-nan1, ZHENG Mao2, ZHENG Lin-hua2 (1. Pingdingshan Industrial College of Technology, Pingdingshan 467001, China; 2. School of Electronic Science and Engineering, National University of Defense Technology, Changsha 410073, China) Abstract: The relationship between Mutual Information principle and the kernel probability density function was analyzed. A novel independent component analysis (ICA) algorithm based on parametric kernel probability density estimation using generalize gaussian distribution was proposed. The kurtosis was calculated by the original data samples and the gaussian exponent was evaluated from the kurtosis by subsection function in the algorithm which could evaluate the score function directly. Based on the score function, the proposed algorithm induced the rules to update the separation matrix by the Mutual Information minimization principle. It solved the important problem of estimating the score function of the sources in ICA to a certain extent. The proposed ICA algorithm was able to separate a wide range of source signals. Simulations show the effectiveness of the proposed algorithm. Key words: independent component analysis; mutual i