Theory Framework of Independent Component Analysis
Ang Y · Microcomputer Information · 2007
Source of the signal- independence in statistical theory is introduced. The mathematics descriptions of signal- independence in higher- order statistics analysis, information theory and parameter estimation theory are deduced. Cost- functions of independence anal- ysis are constructed from the equivalent description, and an extremum expression is introduced to uniform these cost- functions. Inde- pendence decompose methods including gradient methods, fixed- point ICA and Jacobi methods is deduced from the derivative and trigonometric function expression of the extemum expression, which is to solute the cost- function. These works construct the theory framework of the ICA.