Blind separation of co-channel digital signals under imperfect carrier recovery

The separation of co-channel narrowband digital signals is an important problem that is currently being pursued. An important practical corollary to the problem statement and the data model used in these is the fact that the carriers of the different users are only nominally the same but in practice have small differences or offsets. The received signal is actually a sum of digital signals modulated by slightly different sinusoids. We develop a model for an adequately oversampled baseband signal after demodulation with a nominal carrier. We show that by casting certain derived second-order quantities in the format required for application of ESPRIT, all the carrier offsets and the array vectors can be estimated. Further, by suitable weighting and classification of the array outputs the symbol streams of the co-channel users can be separated. Potential applications include uplink algorithms at the base station when the multipath scenario is slowly varying and the delay spread is small