Practical frequency domain ICA algorithm capable of solving permutation and gain ambiguities for digital communication system

Abstract

A computationally efficient frequency domain independent component analysis (ICA) algorithm (ICA-F), which separates the source signals from the measurements of their convolution mixtures, is proposed. Techniques are developed to solve the permutation and gain ambiguities. Simulation results are given to demonstrate the effectiveness of the proposed ICA-F algorithm.