Abstract: Based on the quadratic integral programming model of the multiuser detection problem, a detection strategy by the semidefinite programming method with pretreatment is presented. A pretreatment method is used to obtain an equivalent quadratic integral model with a lesser scale. Then, we give the semidefinite programming relaxation for the equivalent quadratic integral model. Coupled with the randomized method, we obtain the suboptimal solution to the multiuser detection problem. Compared with the semidefinite programming method, our method has the advantages of lower bit error rate and lesser CPU time, which simulations have proved.

Key words: multiuser detection, semidefinite programming relaxation, quadratic integral programming, randomized method, bit error