The Research Of Multiuser Detection Bsased On Wavelet Neural Network

Multiple access interferences (MAI) is the main factor affected the system capacity of the CDMA communication system. But the multi-user detection (MUD) can make fully use of all users' signal information resulting in MAI and have the aimed user to combined detect, therefore, it not only raises the capacity of the system ,but also work out a near-far effect problem ,thus it has good anti-interference performances. Nowadays, a lot of multi-user detection algorithms hold flaws such as sophisticated algorithm, low speed of convergence and so on. As one kind of new optimization method , the artificial neural networks are already applied broadly in recent years, the MUD combine merits of neural network such as quickly speed, stronger processing etc. , becomes a research focus. According to this, algorithms of the MUD based on wavelet neural network are analytically discussed in the paper.The major works of this paper are summarized as follow: 1. Introduced research multi-user detection technical meaning and present condition, and as to performance, characteristics and drawback of the existed main algorithms are summarized and classified. The future development of MUD is discussed at the end .2. In analysis wavelet neural network theory foundation , combine wavelet neural network and multi-user detection, then put forward and discussed MUD based on wavelet neural network. Because of current algorithms of traditional neural network behaved the defects such as longer convergence time,higher bit error rate and so on, this paper discussed wavelet neural MUD algorithms based on back-propagation (BP) and its improved forms. Stimulation shows that the proposed methods are more superior to conventional neural network MUD algorithms in bit error rate and convergence speed.3. Elaborated constant modulus algorithm (CMA) principle , and integrate CMA into wavelet neural network, construct a new cost function, then discuss wavelet neural network MUD based on CMA and constraint CMA. Stimulation indicates that the effectiveness of the proposed approach is more superiority to present algorithms.