Blind Multiuser Detection Based On Neural Network In CDMA Communication System

As one of the key techniques in CDMA communication system, multiuser detection (MUD) can jointly detect the signals transmitted by all users, thereby can reduce multiple access interference (MAI) effectively, relax "near-far" effect dramatically, improve the system performance obviously and increase the system capacity.But the realization of MUD is complex, many research activities have been focused on blind multiuser detection. With the rapid development of neural network, the research of its application in telecommunication is widely carried out. In this paper, for DS/CDMA communication system, by applying neuralnetwork algorithms to blind mulituser detection, we have gained some significative results. The main works of this paper can be summarized as follows:1.Analysed and researched the probability and the theory of using neural network in DS/CDMA communication systems. The simulation results proved the validity and the feasibility of this method.2.Proposed a Lagrange neural network blind multiuser detector in the synchronous CDMA system through white Gaussisn channel, derived its dynamic equations, proved its stability and convergence simply, gave the circuit implement of the neural network. The simulation results showed the proposed algorithm could suppress "near-far" effect, had fast convergence speed, and could be suit to time-varying environment. The simulation results were accorded with the theory analyses.3.In multipath channel, we adopted a new cost function, converted the complex-valued optimization problem to a real-valued one, solved it using Lagrange neural networkefficiently. The simulation results showed the proposed algorithm improved the calculation complexity and the convergence performance.