ICAR: independent component analysis using redundancies
Laurent Albera, Anne Ferréol, Pascal Chevalier, Pierre Comon · 2004
A new blind source separation (BSS) algorithm, called ICAR and using only fourth order (FO) statistics of the data is proposed. The latter method is compared by computer experiments with the well-known methods COM1, COM2, JADE, FastICA, and SOBI. Since ICAR has given very good convergence results and has performed the source separation in the presence of the Gaussian noise with unknown spatial correlation, it appears as being one of the most attracting BSS algorithms.