基于支持向量机的多类数字调制识别方法  

Digital Modulation Recognition Method Based on Support Vector Machines

作　　者：章仁飞[1] 陈印[2] 卢莹 

机构地区：[1]中国电子科技集团公司第38研究所,合肥230031 [2]重庆大学通信与测控中心,重庆400044

出　　处：《世界科技研究与发展》2013年第2期181-183,186,共4页World Sci-Tech R&D

基　　金：中央高校基本科研业务费(CDJZR10160011);重庆市自然科学基金(2010BB2049)资助

摘　　要：提出了一种基于支持向量机的多类数字调制方式识别方法。该方法根据信号的瞬时信息提取特征向量,并利用支持向量机将特征向量映射到一个高维空间,在高维空间中构造最优分类超平面以实现信号分类。该方法避免了判决门限的确定,与传统的神经网络算法相比,具有更好的泛化推广能力。仿真结果表明,在高斯白噪声环境下,信噪比大于等于5 dB时,7种数字调制信号的正确识别率均高于97.4%。A new method for digital modulation recognition algorithm based on Support Vector Machine （SVM） is presented. A set of key features based on instantaneous information of modulation signals were extracted, mapped into a high dimension space. The classification was cartied out in the high dimension space based on SVM. In this method, the decision threshold became unnecessary. And better generalization ability was also acquired comparing with traditional neural networks. Experiments on seven digital modulation signals corrupted by Gaussian noise were conducted. Simulation result shows that the algorithm is quite practical because the overall success rate is more than 97.4% at the SNR of 5 dB.

关 键 词：数字调制识别 支持向量机 瞬时信息 特征提取 

分 类 号：TN911.3[电子电信—通信与信息系统]