Blind signal separation via independent component analysis

Abstract

The “Infomax” method of Independent Component Analysis (ICA) is applied to the problem of separating received signals that overlap in time and frequency, but are otherwise unrelated. The Infomax method separates unknown non-Gaussian signals from a number of signal mixtures by maximizing the entropy of a transformed set of signal mixtures. This work specifically focuses on mixtures of simple communications signals. The Infomax method, as implemented, is found to be successful and efficient for small numbers of signals.