The Research Of Multiuser Detection Based On Particle Filter Algorithms Under Non-gaussian Noise

Mobile communication has profoundly changed people’s lives, as a communicationfrontier the third generation mobile communication system (3G) provide people with a bettercommunication experience.3G technology is based on the theory of spread spectrumcommunication. Its multiply address access project is code division multiple access (CDMA)technology. Compared with the second generation mobile communication, the3G system hasgreater system capacity, security and resistance to fading capability. We can get the differentusers’ data by using the matched filter if the spreading codes of all users are completelyorthogonal to each other. But in practice, the propagation channel is time-varying channel, soit is very difficult to make the spreading codes orthogonal to each other. This leads to mutualinterference between different users’ signals, the interference called multiple accessinterference (MAI). This question severely limits the capacity of the system.Multi-user detection (MUD) technology analyzes multiple access interference, and thenuses all of the users’ information to distinguish from each other, such as spreading codes,amplitude and so on. It can eliminate the correlation between different users’ signals andexpand the system capacity.Firstly, this paper establishes a model of DS-CDMA system, then introduces severalclassic multi-user detection algorithms, including optimal multi-user detection algorithm,decorrelation multiuser detection algorithm and so on. Then, this paper analyzed theadvantages and disadvantages of these algorithms by simulation using MATLAB. Thebackground noise of classic multi-user detection algorithms are assumed to be Gaussian noise,which is reasonable in theory. But in real, Gaussian noise is not existed and most kinds ofnoise have strong pluses. In this environment, the performance of the classic multi-userdetection algorithm will drop significantly. Therefore, a more accurate model is established inthis paper, and Laplace noise and Alpha noise are selected to be background noise.In order to improve system performance and adaptability in non-Gaussian noiseenvironment, firstly, this paper creates a dynamic space model by using whitened matchedfilter. Then the particle filter (PF) algorithm is introduced into multi-user detection, and the core idea and drawback are analyzed in detail. In order to overcome the particle degeneracyof PF algorithm, some new algorithms are proposed, like weight selected particle filter(WSPF) and Gaussian particle filter (GPF). After simulation in MATLAB, we can see that PFand its improved algorithm not only greatly improve the performance of multi-user detectors,but also maintain the robustness in non-Gauss noise.