Research Of Multi-User Detection Technologies For 3G

In spread spectrum CDMA communication technologies developed from the CDMA technology has many advantages, this technology enables system communication capacity, the signal coverage area, call quality excellent, and has a strong resistance to various kinds of interference effects in particular, is confidential in terms of security is very strong and so successful. Communication systems now make up the deficiency, and can be used to meet the many requirements of future communication systems. Therefore, CDMA technology is generally considered the third and fourth generation mobile communication technology first. Of course now, CDMA technology is not perfect, this technique needs further improvement. At present two main defects of this system is inter-symbol interference (ISI) and multiple access interference (MAI). in general, It will appear in near-far problem.CDMA system in order to eliminate MAI, enhance the system's carrying capacity, this thesis to address this situation using the genetic algorithm to train RBF network and the optimized RBF neural network application to multi-user detection. In order to achieve elimination of MAI and ISI, mitigate near-far problem, thus thoroughly improve the system performance, increased system capacity. The RBF neural network is a typical local approximation network, so easy to fall into local minimum, so the need to use genetic algorithm to determine the RBF network structure and the output layer weights. I chose a hybrid genetic algorithm to determine the gradient RBF network, and the least square method to determine the output layer weights. However, the implementation process of the algorithm will appear singular, infeasible solutions, etc., so he used the singular value decomposition method to solve these problems. Ultimate good MAI and ISI elimination of the problems, improve system performance, increase system capacity.