Multiuser detection for DS-CDMA communication systems based on recurrent neural network structures

New results for a multiuser detector based on recurrent neural network structures (MU-RNN) for a direct sequence code division multiple access communication system with multipath propagation are given. Contrary to other neural network approaches the MU-RNN has the advantage, that the network size as well as the weight coefficients of the network can be derived from the parameters characterizing the communication system. The energy function of the MU-RNN matches the loglikelihood function of the maximum likelihood detector and thus has the potential for optimum performance. Different iteration algorithms for the MU-RNN with an emphasis on parallel processing are discussed. The performance and complexity of the MU-RNN are compared with other optimum and suboptimum detection algorithm, specifically the multistage detector.