基于LPC技术的运动车辆目标分类研究  

Classification of Moving Vehicles Based on Linear Predictive Coding

作　　者：丁帅帅 张群[1] 张亮[1] 李开明[1] 王恺[2] 

机构地区：[1]空军工程大学信息与导航学院,陕西西安710077 [2]空军工程大学,陕西西安710077

出　　处：《现代防御技术》2016年第2期177-184,共8页Modern Defence Technology

基　　金：基于图像稀疏分解理论的空间群目标分辨与微动特征提取(61471386);陕西省统筹创新工程-特色产业创新链项目(S2015TDGY0045)

摘　　要：将线性预测编码(LPC)技术应用于运动车辆目标的多普勒特征提取中。首先,基于微多普勒原理建立了车辆目标雷达回波模型,分析了履带式车辆和轮式车辆各自回波的特点。其次,针对杂波干扰问题,采用CLEAN算法对其进行抑制。在此基础上,利用实测数据,提取了不同车辆目标多普勒的LPC系数,给出了预测误差和计算时间在不同LPC阶数条件下的变化趋势。最后利用学习向量量子化(LVQ)方法对提取的LPC系数进行分类,实验结果证明了LPC技术在车辆目标分类中的有效性。Linear predictive coding （LPC） is utilized to the Doppler feature extraction of moving ve- hicles. Firstly, radar returned echoes model of moving vehicles is established based on micro - Doppler effect. The echo characteristics of moving wheeled and tracked vehicle are analyzed. Then, according to clutter jamming, CLEAN algorithm is used to restrain it. On this basis, the coefficients from a linear predictive model are extracted as Doppler features of different vehicles by radar measured date. The chan- ging trends of forecast error and operation time are given with different numbers of LPC coefficients. Fi- nally, learning vector quantity （LVQ） is used for classification. The experimental results validate the ef- fectiveness of LPC method in the classification of vehicles.

关 键 词：线性预测编码 运动车辆 微多普勒 CLEAN算法 学习向量量子化(LVQ) 目标分类 

分 类 号：TN957.51[电子电信—信号与信息处理]