中文摘要：

该文利用机器学习中多核学习（Multiple Kernel Learning, MKL）算法的多源信息融合功能，提出一种基于核层面信息融合的雷达辐射源个体识别框架。对雷达辐射源信号所提取的不同特征表示，分别构建相应的核函数或核矩阵，然后通过一定的准则计算它们的组合系数，并同时或独立获得支持向量机的分类超平面，最终实现对辐射源信号的分类。特别地，该方法能够实现辐射源信号模糊函数多个“近零”切片特征的有效融合，得到比代表性切片更优的识别性能。在3组实测雷达辐射源数据上的实验表明了所提方法的有效性。

英文摘要：

Using Multiple Kernel Learning （MKL） algorithms, which have the function of multi-source information fusion, this paper presents a method for specific radar emitter identification based on kernel-level information fusion. For various feature representations of radar emitter signals, the corresponding kernel functions or kernel matrices are constructed respectively, then their combination coefficients are calculated according to some criteria and the classification hyperplane of Support Vector Machines （SVM） is obtained simultaneously or independently, finally the identification of different emitters is realized. Especially, the proposed methods can effectively fuse the near-zero-doppler slices of Ambiguity Function （AF） of radar signals, getting better performance than the representative-doppler-slice of AF. The experimental results on three real radar data demonstrate the validity of the proposed methods.