Fastica-Based Blind Signal Separation and its Application to Radio Surveillance

This paper presents an efficient blind signal separation scheme for radio surveillance of digital modulation signals using Fast-ICA algorithm. FastICA (Fast Independent Component Analysis) enables very fast optimum weight calculation without any prior information in separating signals by array processing. We first analyze the undesired phase rotation of separated signals caused by FastICA, and then study the compensation technique for the rotation. Moreover, we evaluate signal separation accuracy and computational cost of the proposed algorithm in comparison with conventional algorithms through computer simulation and experiments.