Nonlinear Multiuser Detection based on SVM in Fading Channel with Impulse Noise

In direct sequence code division multiple access system, many users´ signals can be transmitted through the same frequency band. Because of the multiple access interference and the nonlinearity of channel, the performance of linear detector will be degraded substantially. Especially in the case of non-Gaussian noise environment (e.g. impulse noise), traditional linear techniques become incapable. In this paper, support vector machine, a nonlinear method based on the statistical learning theory, is proposed to implement the multiuser detector (MUD). As the simulation results show, the performance of SVM MUD is very close to the optimal detector and prior to the linear MUD. The method of solving the nonlinear classification and regression has huge potential in application of signal processing