We present the application of the fast independent component analysis (fastica) technique for blind component separation to polarized astrophysical emission. We study how the cosmic microwave background (CMB) polarized signal, consisting of E and B modes, can be extracted from maps affected by substantial contamination from diffuse Galactic foreground emission and instrumental noise. We implement Monte Carlo chains varying the CMB and noise realizations in order to asses the average capabilities of the algorithm and their variance. We perform the analysis of all‐sky maps simulated according to the Planck satellite capabilities, modelling the sky signal as a superposition of the CMB and of the existing simulated polarization templates of Galactic synchrotron. Our results indicate that the angular power spectrum of CMB E mode can be recovered on all scales up to l≃ 1000, corresponding to the fourth acoustic oscillation, while the B‐mode power spectrum can be detected, up to its turnover at l≃ 100, if the ratio of tensor to scalar contributions to the temperature quadrupole exceeds 30 per cent. The power spectrum of the cross‐correlation between total intensity and polarization, TE, can be recovered up to l≃ 1200, corresponding to the seventh TE acoustic oscillation.
