Effective blind separation of skewed sources
This article proposes a new algorithm for estimating skewed real signals in blind source separation. It is based on the eigendecomposition of a third-order statistics matrix. This matrix is updated iteratively by the algorithm in order to increase the robustness of the estimated source signals against numerical errors and perturbations. Simulations show the effectiveness of the approach.