论文部分内容阅读
有效的信号特征提取是高精度雷达辐射源识别的基础,以脉冲描述字为代表的传统特征已无法满足复杂电磁环境的需要。本文提出一种基于自适应噪声完备集合经验模态分解(CEEMDAN)的有效雷达辐射源脉内细微特征提取算法。雷达信号由对非平稳、非线性信号尤为有效的CEEMDAN分解产生的个别分量重构,抑噪效果通过1 000次蒙特卡罗实验得到验证,同时设计基于该重构的一种脉内特征空间。本文方法与主流特征提取方法的识别精度在6部雷达辐射源产生的3000个不同脉内调制的加噪信号样本上进行了实验对比,结果表明不同种类信号样本在本文特征空间中清晰可分,本文方法较之主流方法更加精确,尤其在0 d B信噪比(SNR)下仍保持90%以上的高精度。
Effective signal feature extraction is the basis of high-precision radar emitter identification. The traditional features represented by pulse descriptor can not meet the needs of complex electromagnetic environment. In this paper, we propose an algorithm for extracting in-line subtle features of an effective radar emitter based on the complete set empirical mode decomposition (CEEMDAN) of adaptive noise. The radar signal is reconstructed by the individual components resulting from the CEEMDAN decomposition, which is particularly effective for non-stationary, non-linear signals. The noise suppression effect is verified by a 1 000 Monte Carlo experiment and an intra-pulse feature space based on the reconstruction is designed. The accuracy of the proposed method and that of the mainstream feature extraction method are compared experimentally with 3000 different in-pulse modulated noise-added signal samples generated by 6 radar radiators. The results show that different kinds of signal samples are clearly distinguishable in the feature space of this paper, The proposed method is more accurate than the mainstream method, especially at 0 dB signal-to-noise ratio (SNR), maintaining more than 90% accuracy.