Channel Estimation in Unknown Noise: Application of Canonical Correlation Decomposition in Subspaces

The popular subspace algorithm proposed by Mouline et al. performs well when the channel output is corrupted by white noise. However, when the channel noise is correlated as is often encountered in practice, the standard subspace method degrades in performance. In this paper, based on second-order statistics and utilizing Canonical Correlation Decomposition (CCD) to obtain the subspaces, we develop two algorithms to blindly estimate the FIR channels in spatially correlated Gaussian noise with unknown covariance matrix. Our algorithms perform well in any unknown Gaussian noise environment and outperform existing methods proposed for similar scenarios.