An Adaline Neural Network-Based Multi-User Detector Improved by Particle Swarm Optimization in CDMA Systems

The detector applied to each single user based on adaline neural network (ANN) is presented in this paper, which is equivalent to the joint one applied to all users in eliminating the multiple access interference in CDMA systems. Then, particle swarm optimization (PSO) algorithm combined with least mean square scheme is employed in the training procedure of the ANN, which can effectively remove the shortcoming of the poor dynamic adaptive behavior of conventional neural network, i.e. during the convergence procedure of the weights in conventional neural network, the training samples are usually required to be trained iteratively. However, in the improved detector, each training sample can be trained repeatedly, so that the converging speed is getting much faster. Simulation results show that, in the ANN based multi-user detector improved by PSO, the dynamic adaptive behavior has been remarkably improved.