An iterative blind source separation technique: implementation and performance

This paper introduces a source separation technique exploiting the non-Gaussianity of the signals emitted by the sources. The proposed approach relying on non-linear functions shows interesting properties in the noiseless case. Its performance does not depend on the mixing matrix. Closed form expressions of the interference rejection levels are given via an asymptotic performance analysis. These expressions show that the interference rejection levels depend only on the distributions of the source signals and the sample size. This is completed by some numerical simulations showing the important choice of the non-linear functions