基于源信号之间统计独立性的 ICA方法的等价性研究

Study on the Equivalence ICA Methods of the Based on the Statistical Independence

引入了基本ICA问题的发生模型;通过对基于源信号之间统计独立性的三种基本ICA方法(极大似然法、最大信息法和最小互信息法)的推导结果的对比,得出了这三种算法是完全等价的结论;并通过仿真实验,验证了这些算法的有效性。

This paper presents the generation model of independent component analysis(ICA) problem. Three statistics independence based ICA methods, the Maximum likelihood, the Infomax, and the Minimum Mutual Information, are studied by deriving and comparing their iteration equations. The results of the study show that the three methods are totally the same although they are derived from different criterion and cost functions. Simulation for separating a mixed signal is conducted, which verifies the validity of these kinds of ICA methods.