| 基于在线SVM的多用户检测算法及仿真 |
| |
| 引用本文： | 赵宇,奚宏生,王子磊,杨坚.基于在线SVM的多用户检测算法及仿真[J].系统仿真学报,2006,18(1):50-53. |
| |
| 作者姓名： | 赵宇 奚宏生 王子磊 杨坚 |
| |
| 作者单位： | 中国科学技术大学信息科学技术学院自动化系,安徽合肥,230027 |
| |
| 摘 要： | 基于传统支持向量机的多用户检测算法运算量大、耗时久,无法满足实时性要求,为了解决这一问题,结合OSVC方法提出了一种新的多用户检测算法。该算法通过实时增加训练序列,利用KKT条件构造当前训练样本集,实时地调整最优分类超平面。仿真实验表明,该算法所需的支持向量较少,能有效地降低运算负载,抑制了多用户干扰和环境噪声,性能较MMSE多用户检测器有明显的提高。
|
| 关 键 词： | 支持向量机 多用户检测 在线 |
| 文章编号： | 1004-731X（2006）01-0050-04 |
| 收稿时间： | 2004-11-05 |
| 修稿时间： | 2005-08-05 |
Online Training of SVM for Multiuser Detection |
| |
| ZHAO Yu,XI Hong-sheng,WANG Zi-lei,YANG Jian.Online Training of SVM for Multiuser Detection[J].Journal of System Simulation,2006,18(1):50-53. |
| |
| Authors: | ZHAO Yu XI Hong-sheng WANG Zi-lei YANG Jian |
| |
| Institution: | School of Information Science and Technology, USTC, Hefei 230027, China |
| |
| Abstract: | The runtime of conventional SVM-MUD is too long to satisfy the requirement of real-time application.To solve this problem,a new algorithm based on online training of SVM for multiuser detection was proposed.The algorithm added one training sample every step and constructed current training set using KKT condition in order to adjust the maximal margin hyperplane dynamically.Simulation results illustrate that the algorithm has a smaller number of support vectors preserving the same quality of separating hyperplane and shows that it has an excellent effect on multiuser interference and noise suppression.The performance of the online-SVM detectors is much better than that of MMSE detectors. |
| |
| Keywords: | CDMA MMSE |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
|
