Alternate Objective Functions for Independent Component Analysis
P.K. Rajan · Proceedings · 2007
To separate linearly mixed signals which are statistically independent, minimization of objective functions that characterize the independence of the components is employed. Kurtosis, entropy and likelihood functions are some of the functions employed as objective functions. In this paper, directly applying the condition for independence of random signals, alternate objective functions are developed. The suitability of these functions for independent component analysis is investigated.