类相关时频分布在雷达目标识别中的应用

针对雷达一维距离像的非平稳特性,提出一种利用类相关时频分布的高分辨雷达目标识别方法.该方法通过优化时频分布中的核函数来实现在模糊时频平面的特征抽取和特征压缩.在此基础上,提出并实现了利用模拟退火算法同时优化时频特征和识别性能的目标识别方案.仿真实验结果表明,基于类相关时频分布的雷达目标识别方法合理可行,提出的模拟退火核函数优化算法在识别性能上优于已有的Fisher鉴别比优化算法.

Application of class-dependent time-frequency distribution to radar target recognition

Aiming at the non-stationary characteristics of the radar one-dimensional range profile,a new high resolution radar target recognition method utilizing class-dependent time-frequency distribution is presented.Features extraction and features deduction can be implemented in time-frequency ambiguity plane by optimizing the time-frequency kernel function.According to the analysis results,a target recognition scheme which optimizes simultaneously features and recognition capability is introduced and implemented using simulated annealing(SA) algorithm.Simulation experiment results show that the proposed recognition scheme is reasonable and feasible,and the recognition capability derived from the scheme using simulated annealing algorithm outperforms the scheme using Fisher's discriminant ratio algorithm.