基于降维循环谱的特征提取与目标识别

Feature Extraction Based on Reduced-dimensional Cyclic Spectrum and Target Recognition

在日益复杂的电磁环境中,如何提取有效特征是解决目标识别难题的一个关键。通过对目标信号的分析,发现它们虽呈现非平稳性,但却具有循环平稳性。因此,循环谱在分析此类信号方面具有优越的潜力,但是采用循环谱通常导致高维问题。针对这个问题,这里提出了降维循环谱的特征提取与目标识别方法,该方法以循环谱的相同频率点在不同循环频率下的相关性作为识别特征,并用主成分分析方法对该特征降维。实验结果表明,基于降维循环谱的方法具有很好的鲁棒性。

In the increasingly complicated electromagnetic environment,how to extract useful features is a key to achieving target recognition.Target signals have cyclic stationary characteristics and non-stationary characteristics as well,hence cyclic spectrum is superiorly potential in analyzing this type of signals.However,it could induce high dimension problem,and in order to solve this problem,this paper proposes a target recognition method by using reduced-dimensional cyclic spectrum.The method herein is to utilize the correlation of different cyclic frequencies as recognition features.The dimension of the feature space spanned by these features is reduced by Principle Component Analysis（PCA）.Finally,the experimental results show that the proposed method is of good robustness.