A software simulation testbed for multi-user detection based on modified Hopfield network

Several examinations have shown the drawbacks of single user detectors (SUD) and forced a new concept of suitable multiuser detection schemes to increase the efficiency against interference in mobile communication systems. Some suboptimal solution are only available at this moment because of the complexity of implementing the optimal detector. One of this solutions is the recurrent neural network structure. This algorithm shows a high level performance close to the optimal detector. In this paper, the noisy Hopfield network is proposed as MUD structure and simulated through a wireless CDMA multiuser link using Simulink and Matlab. This structure seem to have superior performance compared with the original Hopfield while only a few more iterations is needed.