摘要
采用小波包变换提取雷达辐射源信号特征能够有效对信号进行识别,然而,由小波包变换提取的信号特征维数高,部分信号特征受噪声污染严重.基于此,采用ReliefF算法对信号特征的分类能力进行评价,选择出小波包中分类能力强的信号特征,再通过特征相关度算法去除分类能力相近的冗余特征,利用剩余的分类能力强的信号特征组成特征向量进行分类.仿真实验结果显示,该方法用较少的信号特征能够获得较高的正确识别率.
Features extracted by wavelet packet transform can effectively identify signals, however, feature dimension of the features is high and some characteristics are seriously polluted by the noise. Classification ability of the features was evaluated by ReliefF algorithm and features with strong classification ability in wavelet packets were picked out. Then redundant features of similar classification ability were deleted by characteristic similarity algorithm. The remaining features of strong classification ability composed eigenvectors and they would be classified. Simulation experimental results show that this method using fewer features can obtain higher right recognition rate.
出处
《成都大学学报(自然科学版)》
2012年第2期151-153,共3页
Journal of Chengdu University(Natural Science Edition)