Improved multiuser blind CCD-based vector channel estimation in colored noise

In this paper, we propose and study a generalized criterion for the canonical correlation decomposition (CCD) based estimation of multiple vector channels in ambient colored noise. In this approach, it is assumed that the desired signals are received by two separated antenna clusters, so that the output noise is spatially uncorrelated. The proposed criterion exploits multiple signal codes through a weighted sum of projection errors, which incorporate the kernel matrices of the signals sharing the same target channel. Through computer simulations, we study the effects of using multiple codes in the blind CCD-based channel estimation and we show that the new criterion may indeed lead to quite significant performance improvements in the estimation of the multipath vector channels.