摘要
由于多载波传输数据和基于快速傅里叶变换(Fast Fourier Transform,FFT)的矩形窗截取,MC-CDMA(Multi-Carrier CDMA)系统对频率偏差十分敏感,同时上行链路信号的异步性和信道的非线性,令MC-CDMA从单载波扩频系统继承来的线性多用户检测手段和分集合并技术变得不再有效,信道估计困难而不准确。为了在这种条件下提高信号检测的性能,提出了基于支撑向量机增量学习算法(Incremental Support Vector Machine,ISVM)的MC-CDMA上行链路多用户检测器。该检测器在标准支撑向量机(Support Vector Machine,SVM)的基础上舍弃历史样本,减少不必要的训练,同时合理地处理了新增样本和原支撑向量机分界面的关系,保留了强大的非线性分类力。通过仿真实验,与常用的检测技术以及最佳检测曲线比较,表明该检测器能很好地逼近最佳检测器。
An uplink Multi-Carrier CDMA（MC-CDMA） multi-user detector based on Incremental Support Vector Machine（ISVM） is proposed in this paper. The proposed detector minishes training set by discarding history samples and maintains strong ability of nonlinear classification with more proper incremental learning methods based on reasonable analysis of relationship between new training samples and old Support Vector Machine（SVM）＇s hyperplane. Comparison curves are drawn between the proposed detector and traditional detection technologies, and simulation results show that the proposed detector based on ISVM can well approach the optimum detection performance.
出处
《信息与电子工程》
2006年第6期421-426,共6页
information and electronic engineering
