摘要
研究脑电图成像的数据处理问题时,独立成分分析(ICA)是一种新的信号处理统计方法,被广泛用于各个领域。脑电图就是,利用独立成分分析从混合信号中还原出源信号,通过目标函数,如极大似然估计,信息最大化和互信息最小化等,对源信号的概率密度函数(PDF)进行估计。在基于互信息最小化算法的基础上,提出一种新的独立成分分析算法,算法中的核心参数是由信号本身来确定的,能使所估计的PDF更加准确,从而提高分离的性能。最后,用新的ICA算法来实现脑电图(EEG)信号的盲源分离,结果表明,算法可以快速有效的分离其源信号,且准确性优于Boscolo提出的非参量ICA模型。
Independent component analysis（ICA） is a new method of signal statistical processing and widely used in many fields.To separate source signals from mixed signals by ICA,probability density function（PDF） of source signals through some object functions,such as maximum likelihood estimation,information maximization and mutual information minimization,should be estimated.So a new ICA algorithm based on mutual information minimization is generated,and the kernel parameters in the algorithm are determined by signals themselves.It can make the PDF of signals more accurate,and then improve the separation performance.In the end,the blind sources separation of real electroencephalogram（EEG） signals is realized.The result shows that the new algorithm can separate the source signals of EEG quickly and effectively and the accuracy is higher than Non-Parametric ICA algorithm proposed by Boscolo.
出处
《计算机仿真》
CSCD
北大核心
2010年第11期217-220,共4页
Computer Simulation
关键词
独立成分分析
互信息熵
概率密度函数
脑电图
Independent component analysis（ICA）
Mutual information entropy
Probability density function（PDF）
Electroencephalogram
