Abstract: A nonlinear multiuser detector based on the support vector machine(SVM) is presented. The presented detector can approach the optimum multiuser detector by using the Midamble of TD-SCDMA. Structural risk minimization(SRM) introduced in SVM leads to the greater generalization and fewer training samples than the other learning models. The nonlinear support vector machine multiuser detector can approach the optimum multiuser detector in a better way than the other linear detectors. The resuls of simulation show the feasibility and validity of this detector in the multipath channel.

Key words: TD-SCDMA, multiuser detection, support vector machine