基于EMD和能量比的战场声目标分类与识别  

BATTLEFIELD ACOUSTIC TARGET CLASSIFICATION AND RECOGNITION BASED ON EMD AND POWER RATIO

作　　者：吕艳新[1] 孙书学[1] 顾晓辉[1] 

机构地区：[1]南京理工大学机械工程学院,南京210094

出　　处：《振动与冲击》2008年第11期51-55,61,共6页Journal of Vibration and Shock

摘　　要：采用线性预测方法对信号进行边界延拓,改进EMD方法,应用EMD(经验模态分解)对战场声信号进行分解,对分解得到的有限个IMF(本征模态函数)进行FFT,求得其相应的幅值谱,进而得到其能量。选择每一个IMF的能量相对于原始信号总能量的能量比作为特征向量,并将其归一化。最后,设计神经网络分类器对不同类战场声目标进行分类与识别。实验结果表明,基于EMD和能量比的战场声目标分类与识别,分类效果显著,识别率较高。The ends of acoustic target signal were extended by linear prediction method.The signal was decom- posed into finite IMFs by the improved EMD(Empirical Mode Decomposition).The frequency spectrum and the power of each IMF(Intrinsic Mode Funetion) were calculated by using FFT.The unitary ratios of power of each IMF to that of orig- inal acoustic signal were chosen as the components of eigenvector.Then battlefield acoustic targets were classified and rec- ognized by neural network designed.The results show that the classification effect based on EMD and power ratio in battle- field is remarkable and the recognition rate is rather high.

关 键 词：经验模式分解 能量比 特征提取 神经网络 分类与识别 

分 类 号：TN911.72[电子电信—通信与信息系统]