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

As mobile technology has developed from 1G to 3G, it has changed totally. And now the main technology of 3G is CDMA, which has more advantages than the former for its bigger capacity, better safety and lower transmitting power. It is known by the theories that you can get the figures you want just by assessing the output of the matched filter with a traditional detector, if the spread codes of all users connecting to the system are complementary. But in fact, due to the relativity between the spread codes, there exist some interference among the users, which we call Multiple-Access –Interference(MAI). And it has become the key to limit the performance of CDMA.After learning the basic theories of how MAI forms, we introduce Multi-User Detector(MUD). In this technology, all users’ information gets involved in detecting. Besides, it can get rid of MAI between different users and improve the performance of the system.Initially based on the model of multi-users’ system, the subject tells us the basic theory of detecting technology and the composition and function of detecting system. It also introduces some typical multi-user detectors of linear and nonlinear to improve its functions. This typical multi-user detector often supposes that the noise of background is Gaussian Noise, which is likely to be true. On the contrary, when in non-gaussian noise, it works a little bad. Because of this, we put forward particle filter algorithm which has advantages of a flexible application, a wide range of application and a high precise detection. It can be used in all stochastic systems, especially in non-gaussian filter. Meanwhile, combined with CDMA, this algorithm can be introduced into the detection of multi-user. With the emulation of MATLAB, it can be better than other typical MUD.Because the particle can deteriorate, which can influence the performance of detecting, we get two advanced versions of algorithm—Gauss- Hermit particle filter algorithm and regularized particle filtering algorithm. When in the same situation, as the experiment goes, this two new are better.