Self-organizing multiuser detection

The conventional DS/CDMA receiver utilizes a code matched filter followed by a sign decision. In a multiuser environment this approach is suboptimal and it does not lead to a near-far resistant (NFR) receiver. A NFR receiver, optimum or suboptimum, utilizes the multivariate statistics provided by the bank of matched filters in order to make a decision for any single user. Several multiuser detection algorithms have been proposed. This paper introduces an adaptive CDMA multiuser detector. The proposed detector combines channel estimation and data detection in a recursive structure driven by learning rules motivated by a self-organizing neural network. Both data-aided and blind learning are possible at a reasonable computational cost. The performance of the self-organizing detector is compared with other multiuser detection schemes and simulation results are provided