Improving signal subspace estimation for blind source separation in the context of spatially correlated noises

We address the issue of orthogonal techniques for blind source separation of periodic signals when the mixtures are corrupted with spatially correlated noises. The noise covariance matrix is assumed to be unknown. This problem is of major interest with experimental signals. We first recall that principal components analysis (PCA), cannot provide a correct estimate of the signal subspace when the noises are spatially correlated or when their power spectral densities are different. We then introduce a new estimator of the non-noisy spectral matrix using delayed blocks. The only assumption is that the noise correlation and cross correlation lengths must be shorter than the source correlation lengths. Simulation results show the efficiency of the new method.