摘要
利用2个高阶累积量之比构造的特征参数代表的信号特征通常有限,基于此类特征参数的调制样式识别性能有待进一步提高。为此,提出多个高阶累积量组合的调制样式识别算法。采用2个累积量和/差的归一化来构造特征参数,充分利用多个高阶累积量信息,使之包含更多的信号特性;并使用随机森林作为分类器,克服弱分类器存在的过拟合问题。仿真结果表明:提出的特征参数对调制样式的识别性能优于归一化累积量的特征参数,随机森林分类性能优于弱分类器。
The feature parameter constructed by using the ratio of two higher-order cumulants generally represents a limited signal feature,the modulation type recognition performance based on such feature parameters needs to be further improved.In this paper,we propose a modulation type recognition algorithm based on of combination of multiple high-order cumulants.The normalized sum and difference of two cumulants are used to construct the feature parameters to make full use of multiple high-order cumulant information,which include more signal feature.A random forest is used as a classifier to overcome the over-fitting of weak classifiers problem.The simulation results show that the proposed feature parameters are better than the normalized cumulant feature parameters,and the random forest classification performance is better than the weak classifier.
作者
翁建新
赵知劲
占锦敏
WENG Jianxin;ZHAO Zhijin;ZHAN Jinmin(School of Communication Engineering,Hangzhou Dianzi University,Hangzhou Zhejiang 310018,China;State Key Lab of Information Control Technology in Communication System,The 36th Research Institute of China Electronics Technology Group Corporation,Jiaxing Zhejiang 314001,China)
关键词
特征参数
分类器
高阶累积量
随机森林
feature parameters
classifier
high-order cumulant
random forest
