Study Of Multi-user Detection Technology Based On Neural Network

The CDMA multi-user detection is one of the main technologies of the third generation communication system. In CDMA communication system, there always have some relativity in the sequence assigned to the users, which is the root of multiple access interference. Multi-user detection (MUD) makes joint detection on the received signals by making full use of the information of all users' based on tradition detection technology. So it has good anti-interference capability, can solve far and near effect problem, reduce demand for power control precision, notability advance capability of whole system.This paper introduces the principle of the optimal MUD and some classical suboptimum MUD based on the research on the model of CDMA communication system and the principle of MUD. It introduces in detail the detection based on Hopfield neural network (HNN), which translates the problem of minimizing the objective function of the optimal MUD into resolving the minimization of energy function for HNN and has the disadvantage of easily getting into the local minimization of energy function. Aim at this disadvantage, this paper gives a model of hybrid MUD based on reduced detection and the stochastic HNN with soft decision feedback. The reduced detection of hybrid detection provides a better initial value to HNN, then HNN will not get into the local minimization of energy function and get the global optimal solution to improve detection performance. The simulation results indicate that hybrid MUD has better performance in anti-multiple access interference and far and near resistance, and has a good convergence.This paper researches the structure and elements of chaos neural network. It analyzes the structure and dynamics of chaos neural network and transient chaos neural network proposed by Chen and Aihara. The transient chaos neural network has more abundant dynamics characteristic neural network than the Hopfield neural network. It can advance the network's searching ability of global minimum efficiencily.This paper optimizes transient chaos neural network with simulated annealing mechanics. It can accelerate the search speed and guarantee the assurance of the veracity of the optimal arithmetic. The improved transient chaos neural network is researched and applied to the MUD. Experiment proves that this MUD has good performance and there has some theoretic meaning and practicality value in it.