摘要
为快速识别直升机与坦克的声信号,提出了一种变分模态分解(VMD)与分形维数相结合的声学目标识别方法。对目标声信号进行VMD分解,根据目标声信号的频域能量分布确定分解的模态(IMF)数目;通过数学形态学方法计算各个IMF的容量维数与信息维数,并以此为特征向量;利用BP神经网络分类器对直升机与坦克声目标进行识别,通过与传统EMD融合计盒维数的方法进行对比,验证了所提方法具有更高的识别率和运算速度。
In order to quickly identify different acoustic target,signals of helicopters and tanks the Variational Mode Decomposition(VMD) combined with fractal dimensions is proposed. According to the frequency domain energy distribution of the two kinds of acoustic signals,the number of IMFs decomposed by VMD is determined. And then,calculate the capacity dimension and the information dimension of each IMFs of the two kinds of sounds by the mathematical morphologic method. The fractal dimensions are used as the feature vectors by the BP neural network classifier to recognize the acoustic targets. Compared with EMD and the means of box counting,the proposed method shows it has a higher recognition rate and calculation speed.
作者
张坤
顾晓辉
邸忆
ZHANG Kun;GU Xiao-hui;DI Yi(Nanjing University of Science and Technology,Nanjing 210094,China)
出处
《火力与指挥控制》
CSCD
北大核心
2019年第8期77-81,88,共6页
Fire Control & Command Control
基金
086国家科技重大专项基金资助项目（004040204）
