Blind MMSE-constrained multiuser detection

In this paper, blind multiuser detection is studied for a direct-sequence (DS) code-division multiple access (CDMA) system with unknown multipath channels. The receiver is designed to follow a minimum mean-square-error (MMSE) form parameterized by a channel-like constraint vector, which is then optimized by the conventional constant modulus algorithm in a much smaller dimensional space. It is analytically established that the constraint vector asymptotically converges to the channel vector of the user of interest. Correspondingly, the receiver converges to the desired user's MMSE receiver, ensuring detection of the desired signals. When the constraint vector is treated as a channel estimate, the performance of the channel estimator is analyzed using a perturbation technique. Simulation results show a satisfactory performance of the proposed approach, which is also observed to be superior to other approaches in many scenarios.