中文摘要：

高分辨距离像（high resolution range profile,HRRP）是目标沿雷达视线方向上的一维压缩投影,传统的HRRP目标识别方法大都利用单次HRRP测试样本判决。但是,由于单次测试样本包含的信息有限,且容易受到噪声污染,识别鲁棒性难以保证。提出1种基于子空间的HRRP序列噪声稳健识别算法。该算法在训练、测试阶段均利用HRRP序列,基于主成分分析（principal component analysis,PCA）方法生成能够抑制噪声、冗余分量的目标信号子空间,并根据Grassmann流形定义子空间距离,将测试子空间与训练子空间按照最小子空间距离的准则作匹配比较,从而判定测试样本序列所属类别。文章推导证明了传统的最小重构误差方法是提出方法只使用单次HRRP测试样本的特殊情况。基于实测数据的识别实验显示,由于更充分地利用了HRRP序列信息且子空间能够抑制噪声,提出方法较最小重构误差方法具有更好的识别性能和噪声稳健性。

英文摘要：

High resolution range profile（HRRP）is one-dimensional projection of the target echo onto the radar line of sight（RLOS）.Most traditional recognition methods for HRRPs are based on a single HRRP test sample.However,since the single test sample just contains limited information and is easy to be contaminated by noise,it is difficult to guarantee the robustness of these methods.In this paper,a robust recognition algorithm for HRRP sequences based on subspace is proposed.The algorithm utilizes HRRP sequences in both training and test stages.The target signal subspaces generated via Principal Component Analysis（PCA）method can eliminate the noise and redundant components,and the distance between the subspaces is defined according to the Grassmann manifold.The category of test sample sequence is determined by the comparison between training and test subspaces based on the criterion of the minimum distance.In this paper,it is proved that the traditional method based on the minimum reconstruction error is a special case of the proposed method by using only a single HRRP test sample.Experimental results on the measured data show that the proposed algorithm has better recognition performance and is more robust to the test noise,because of more fully using the HRRP sequence information and suppressing noise via subspaces,compared with the minimum reconstruction error method.