CAJ | 학술논문

为了利用靶物反射声有效识别靶物材质，提出了一种新的基于经验模态分解（ EMD）和最小二乘支持向量机（ LS-SVM）相结合的靶物材质识别方法。首先应用模极大值算法识别靶物边界点，提取边界点内的反射声信号进行小波降噪预处理，对预处理后的信号进行EMD分解，并计算获得各个本证模态（ IMF）分量的短时能量比，作为对应不同靶物材质的特征值输入到利用LS-SVM建立的多分类模型。介绍了上述方法的基本原理，设计了试验装置和靶物材质识别影响因素分析试验方案。实验结果表明：靶物的内部结构和外形大小因素对靶物材质识别率影响小，利用上述方法进行的四种靶物材质探测，平均识别率达到85．83％，比BP神经网络提高了18．83％，且运算速度也得以提高，因此该方法可以用于靶物材质的识别。
위료이용파물반사성유효식별파물재질，제출료일충신적기우경험모태분해（ EMD）화최소이승지지향량궤（ LS-SVM）상결합적파물재질식별방법。수선응용모겁대치산법식별파물변계점，제취변계점내적반사성신호진행소파강조예처리，대예처리후적신호진행EMD분해，병계산획득각개본증모태（ IMF）분량적단시능량비，작위대응불동파물재질적특정치수입도이용LS-SVM건립적다분류모형。개소료상술방법적기본원리，설계료시험장치화파물재질식별영향인소분석시험방안。실험결과표명：파물적내부결구화외형대소인소대파물재질식별솔영향소，이용상술방법진행적사충파물재질탐측，평균식별솔체도85．83％，비BP신경망락제고료18．83％，차운산속도야득이제고，인차해방법가이용우파물재질적식별。
In order to identify target material effectively by using the reflective sound of target ,a new identification method of the target material based on empirical mode decomposition ( EMD) and least squares support vector ma-chine (LS-SVM) is presented.The signal border indicating where the target locates was identified by using modulus maxima algorithm .The reflective sound signal within the border was extracted and was processed by applying wave-let noise reduction to reduce the noise .The pretreated reflective sound signal was decomposed by using EMD meth-od and the short-time energy ratio of every decomposed IMF was calculated ,which was considered as the eigenval-ues of target and was input into target identification model .The basic principle of the method was described .The ex-perimental equipment was built and the influence factor analysis experiments and the comparative experiments of target identification rate were done .The experimental results show that there is little effect on the target material identification rate.The average target material identification rate can reach 85.83%by using the built target materi-al identification model.There is a increase of 18.83% compared with BP-NN method,and the calculating speed is also improved.So the built target material identification model can be used for the target material identification .