A support vector machine （SVM） based multiuser detection （MUD） scheme in code-division multi- ple-access （CDMA） system is proposed. In this scheme, the equivalent support vector （SV） is obtained through a kern...A support vector machine （SVM） based multiuser detection （MUD） scheme in code-division multi- ple-access （CDMA） system is proposed. In this scheme, the equivalent support vector （SV） is obtained through a kernel sparsity approximation algorithm, which avoids the conventional costly quadratic pro-gramming （QP） procedure in SVM. Besides, the coefficient of the SV is attained through the solution to a generalized eigenproblem. Simulation results show that the proposed scheme has almost the same bit er-ror rate （BER） as the standard SVM and is better than minimum mean square error （MMSE） scheme. Meanwhile, it has a low comoutation complexity.展开更多
In order to suppress the multiple access interference (MAI) in 3G, which limits the capacity of a CDMA communication system, a fast relevance vector machine (FRVM) is employed in the multiuser detection (MUD) scheme. ...In order to suppress the multiple access interference (MAI) in 3G, which limits the capacity of a CDMA communication system, a fast relevance vector machine (FRVM) is employed in the multiuser detection (MUD) scheme. This method aims to overcome the shortcomings of many ordinary support vector machine (SVM) based MUD schemes, such as the long training time and the inaccuracy of the decision data, and enhance the performance of a CDMA communication system. Computer simulation results demonstrate that the proposed FRVM based multiuser detection has lower bit error rate, costs short training time, needs fewer kernel functions and possesses better near-far resistance.展开更多
文摘A support vector machine （SVM） based multiuser detection （MUD） scheme in code-division multi- ple-access （CDMA） system is proposed. In this scheme, the equivalent support vector （SV） is obtained through a kernel sparsity approximation algorithm, which avoids the conventional costly quadratic pro-gramming （QP） procedure in SVM. Besides, the coefficient of the SV is attained through the solution to a generalized eigenproblem. Simulation results show that the proposed scheme has almost the same bit er-ror rate （BER） as the standard SVM and is better than minimum mean square error （MMSE） scheme. Meanwhile, it has a low comoutation complexity.
文摘In order to suppress the multiple access interference (MAI) in 3G, which limits the capacity of a CDMA communication system, a fast relevance vector machine (FRVM) is employed in the multiuser detection (MUD) scheme. This method aims to overcome the shortcomings of many ordinary support vector machine (SVM) based MUD schemes, such as the long training time and the inaccuracy of the decision data, and enhance the performance of a CDMA communication system. Computer simulation results demonstrate that the proposed FRVM based multiuser detection has lower bit error rate, costs short training time, needs fewer kernel functions and possesses better near-far resistance.
