Blind receiver based on instantaneous independent component analysis for convolutive MIMO systems

In this paper a method is proposed that significantly enhances the previous work on subspace based channel shortening for the blind separation of convolutive MIMO systems. In the subspace based channel shortening, the convolutive system is transformed into instantaneous system thorough successive deflations and can end up to different memoryless systems corresponding to nonzero taps of MIMO channel. However, only one of the memoryless systems is exploited to separate the mixtures by the independent component analysis (ICA) method. In this system we improved the performance by taking advantage of using all the memoryless systems in the estimation of the mixtures. In fact the mixtures are estimated separately based on each evaluated memoryless system and then average is taken over their results. The simulation results show considerable enhancement in performance with a price which is paid for complexity.