一种快速的在线SVM多用户检测算法  

A Fast Online SVM Algorithm for Multi-user Detection

作　　者：王子磊[1] 奚宏生[1] 赵宇[1] 盛延敏[1] 

机构地区：[1]中国科学技术大学信息科学技术学院自动化系,合肥230027

出　　处：《电子与信息学报》2006年第8期1386-1390,共5页Journal of Electronics & Information Technology

基　　金：国家863计划(2003AA103710)资助课题

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。该文提出了一种快速的在线支持向量机多用户检测算法。该算法利用KKT条件判别实时增加的训练序列并构造当前训练样本集,从而能够有效地减少训练样本大小,加快训练速度。仿真实验表明,该算法在不影响分类效果的情况下大大加快了训练速度,且用于分类的支持向量较少,同时性能与传统支持向量机算法相当且明显优于MMSE(RLS)多用户检测器。The runtime of conventional SVM-MUD is too long to satisfy the requirement of real-time application. A fast algorithm based on online training of SVM （FOSVC） for multiuser detection is proposed in the paper. The algorithm distinguishes new added samples and constructs the current training data set using KKT condition in order to reduce the size of training samples. As a result, the training speed is effectively increased. Simulation results illustrate that the algorithm has a faster training speed and a smaller number of support vectors preserving the same quality of separating hyperplane. The performance of the FOSVC detectors is pretty much the same thing as that of SVM detectors, and much better than that of MMSE detectors.

关 键 词：CDMA 在线支持向量机 多用户检测 快速算法 

分 类 号：TN914.5[电子电信—通信与信息系统] TN911.23[电子电信—信息与通信工程]