采用双向LSTM模型的雷达HRRP目标识别  

Radar HRRP target recognition by the bidirectional LSTM model

作　　者：徐彬[1,2] 陈渤[1,2] 刘家麒 王鹏辉[1,2] 刘宏伟[1,2] XU Bin;CHEN Bo;LIU Jiaqi;WANG Penghui;LIU Hongwei(National Key Lab. of Radar Signal Processing, Xidian Univ., Xi'an 710071, China;Collaborative Innovation Center of Information Sensing and Understanding, Xidian Univ., Xi'an 710071, China)

机构地区：[1]西安电子科技大学雷达信号处理国家重点实验室,陕西西安710071 [2]西安电子科技大学信息感知技术协同创新中心,陕西西安710071

出　　处：《西安电子科技大学学报》2019年第2期29-34,共6页Journal of Xidian University

基　　金：国家自然科学基金(61771361;61701379);国家杰出青年科学基金(61525105)

摘　　要：针对传统雷达高分辨距离像目标识别方法只考虑高分辨距离像样本的包络信息而没有考虑其时序相关特性问题,提出了一种采用双向长短时记忆网络模型的识别算法。该算法首先选取出高分辨距离像样本的目标区域,并根据目标区域提取对于平移敏感性稳健的输入特征;然后采用双向长短时记忆模型对输入特征提取双向时序信息;最后通过投票策略融合双向信息,输出样本类别。采用实测数据的实验结果表明,该算法不仅可以有效地识别雷达目标,而且对于平移敏感性非常稳健。As the traditional recognition methods utilize the envelope information of the high resolution range profile(HRRP)without considering the temporal correlation between range cells,a novel recognition method based on the bidirectional long short term memory(BLSTM)is proposed.First,the method extracts the time-shift robust input representation by figuring out the target areas.Then,to consider the bidirectional correlation between range cells,a bidirectional long short term memory is utilized to extract the bidirectional information from the input representations.Finally,the model uses a voting mechanism to combine the bidirectional information and predict the labels.Experimental results based on measured data show that the proposed model is efficient in recognition and robust to the time-shift sensitivity.

关 键 词：雷达自动目标识别 长短时记忆模型 高分辨距离像 平移敏感性 时序相关性 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TN959.17[自动化与计算机技术—控制科学与工程]