Research On Technologies Of CDMA Multi-user Detection

In CDMA communication system the cross correlations among signals of users result in the Muitiple Access Interference (MAI), which is low when the number of users in the same channel is small, but will increases and become the main jamming of broad CDMA communication system with the increase of the number of users or the difference of user's power. The conventional single user detector matches the Spread-Spectrum codes for each user completely according to the classic theory of Directe-Sequences Spread -Spectrum (DS-SS) theory, so its ability to resist MAI is poor.As one of the most important technologies of CDMA, Multi-user Detection (MUD) technique has a good performance of resisting MAI and near-far effect through the use of adequate cross correlations among spreading codes of users to detect the signals.The Optimum Multiuser Detector proposed by Verdu can solve the problem of MAI and near-far effect of CDMA communication system. But the complexity of the detector appears to be exponential in the number of users. When the number of users is large, the detector can't be realized. So high efficiency sub-optimum algorithms of sub-Optimum Multiuser Detectors are important to the researching on MUD.In this dissertation the author proposed a kind of sub-optimum algorithm, a hybrid detector based on Stochastic Hopfield Neural Network (SHNN) and Reduced Detector (RD) for CDMA.The similarity between the energy function of Discrete Hopfield Neural Network (DHNN) and the formula of OMD means that using DHNN to solve the problem of OMD is feasible. The convergence process of Neural Network is the process to solve the problem of OMD. So a MUD based on DHNN has been proposed in this dissertation. Considering that the DHNN may converge to the local minima, the author introduced the SHNN to make the network converge to the global minima. In order to decrease the searching space and complexity of SHNN, to detect the users with higher power, and to make use of the merits of SHNN, a Reduced Detector has been introduced. The main merit of Reduced Detecor is far-near resistance. In order to use the merits of these two detectors synthetically, the hybrid detector has been proposed, which uses the part of Reduced Detector to pretreat the data and the part of SHNN to search the most likelihood sent datas. The simulated results showed the good performance of the hybrid detector.The algorithm is digital scheme, which can be validated and improved in the futurewith the development of DSP technology.