提出了一种改进的ReliefF算法,并将其用于雷达高分辨距离像(high resolution range profile,HRRP)目标识别。与传统ReliefF算法相比,新算法通过在每类目标中等距离间隔抽取相同数量样本的方式进行权值累积,降低了样本数量及分布差异等因素对特征权值的影响,得到了更稳定有效的特征权值。利用此权值不但可降低特征向量维数,并可对最小距离分类器加权,提高目标识别率。最后,对5种不同飞机实测数据的识别结果表明本算法可达到83%的平均识别率。
The existing directions-of-arrival （DOAs） estimation methods for two-dimensional （2D） coherently distributed sources need one- or two-dimensional search, and the computational complexities of them are high. In addition, most of them are designed for special angular signal distribution functions. As a result, their performances will degenerate when deal with different sources with different angular signal distribution functions or unknown angular signal distribution functions. In this paper, a low-complexity decoupled DOAs estimation method without searching using two parallel uniform linear arrays （ULAs） is proposed for coherently distributed sources, as well as a novel parameter matching method. It can resolve the problems mentioned above efficiently. Simulation results validate the effectiveness of our approach.
