摘要
提出一种基于相像系数(RC)的特征选择新方法,给出了RC的定义和基于RC的类别可分离性判据,描述了 基于RC和量子遗传算法的雷达辐射源信号特征选择算法,设计了神经网络分类器,并将该方法与基于距离准则的顺序前 进法(SFSDC)和吕铁军的方法(GADC)作了特征选择和分类识别的对比实验。结果表明,本文方法无需事先指定最优特征 子集的维数,能可靠有效地选择出最佳特征子集,不仅大大降低了特征向量的维数,简化了分类器的设计,而且获得了比 原始特征集、SFSDC和GADC更高的正确识别率和识别效率。
A novel approach called resemblance coefficient (RC) feature selection algorithm was proposed. Definition, properties and the evaluation criterion of the optimal feature subset were given. Feature selection algorithm using RC and quantum genetic algorithm was described and neural network classifiers were designed. The introduced approach, sequential forward selection using distance criterion (SFSDC) and the method presented by Tiejun L ii (GADC) Were used respectively to select the optimal feature subset and to recognize radar emitter signals. Simulation experiment results show that the proposed approach not only can select the optimal feature subset reliably and effectively without designating the prior dimension of the feature subset to lower the dimension of feature vector greatly and to simplify the classifier design, but also achieves much higher accurate recognition rates than original feature set, SFS-DC and GA-DC, respectively.
出处
《信号处理》
CSCD
北大核心
2005年第6期663-667,共5页
Journal of Signal Processing
基金
NEWA251435
QT220401国家自然科学基金资助项目(No.60572143)
关键词
信号处理
特征选择
相像系数
雷达辐射源
pattern recognition
feature selection
resemblance coefficient
radar emitter