The Research Of Multiuser Detection Based On Em Accelerate Algorithms Under Non-Gauss Noise

As the third generation mobile communication system of Multiple Access plan，CDMA is based on the orthogonal Code,and the relevant receiving the theoretical, tospread spectrum communication technology as the foundation of multi-accesstechnologies, high capacity, fight multipath fading capability, high efficiency and easyto seamless frequency switching, good secrecy, etc. According to CDMA, if spreadcodes between users are completely orthogonal, simple code match filters can be usedto distinguish the data of multiusers.However, due to the random access fromdifferent users, it is hard to realize the orthogonality of spread codes andmultiple-access interference, MAI, arise correspondingly. multi-user detection aimsworks as the combined detection of multiusers using other interference.Under optimalconditions, by virtue of multi-user detection the system performance can act as well asone single user involved does. Because of the elimination of MAI, Users can achievereliable performance under the smaller SNR; the reduction in SNR then turns into theincrease of system capacity. Since it has the ability of improving spectrum efficiencyand system capacitance, multi-user detection has become a critical part in mobilecommunication and a hotspot for research.In order to simplify the mathematical analysis chosen on the assumption noise isGaussian while we analyze the system, this assumption is in line with guidelines forthe central limit theorem, not only are these detection in noisy environments are easyto implement, but also indeed demonstrated that in many practical problems thisassumption is appropriate. However, in some natural and artificial resulting from theimpact of impulsive ambient noise case, such as thunder and lightning, ice avalanches,all kinds of machine motors, neon signs, etc. do not have Gaussian nature. This kindof noise shows significant peak amplitude, or "outliers" in the time domain. In orderto prevent the detection performance decline under the "noise spikes". Establish amodel that is more accurate than the Gaussian model is very necessary. In this paper,the Non-Gaussian Noise is considered fully, and we choose DS-CDMA as thecommunication system model. Firstly, several classical multi-user detection areintroduced, such as optimal detection algorithm, linear detection algorithm, andnon-linear detection algorithm; then, analysis the various algorithm thought, and summarizes their performance. In order to improve the detection performance,introduce the EM algorithm which is a simpler calculation, and performance isapproaching to the optimal detection, and EM algorithm has good convergence andperfect robust, but the convergence rate is shortcomings of EM algorithm itself whichaffects its application. Inoder to improve the convergence speed, some new acceleratealgorithms, Aitken-EM, Quasi-Newton-EM, Aitken2EM and Vector-ε EM, areproposed. The simulation results show that the improved algorithm has fasterconvergence speed, satisfy the requirements of real-time processing communicationsystem.