Recursive blind source separation for BPSK signals

This paper addresses the important problem of separating and estimating the contributions of multiple users (sources) signals impinging on an antenna array in a wireless communication system. The proposed approach has key advantages over existing ones, which are listed below: (i) the structure of estimation is recursive, therefore it estimates the incoming bits on the fly as they reach the receive array instead of estimating a separating matrix. (ii) it exploits the finite alphabet property (here, BPSK) of the communication signals in an explicit manner, through the use of a finite-state trellis. (iii) perfect decisions are obtained in the absence of noise upon certain mild conditions on the transmitted signals (excitation conditions). Therefore (local) convergence is not an issue here