无人机飞行声特征与提取方法比较  

Comparison of UAV flight acoustic characteristic and its extraction method

作　　者：金恒康 张一闻 王耀杰 JIN Hengkang;ZHANG Yiwen;WANG Yaojie(College of Information Engineering,Engineering University of PAP,Xi’an 710086,China;College of Cryptographic Engineering,Engineering University of PAP,Xi’an 710086,China)

机构地区：[1]武警工程大学信息工程学院,陕西西安710086 [2]武警工程大学密码工程学院,陕西西安710086

出　　处：《现代电子技术》2019年第22期103-107,112,共6页Modern Electronics Technique

基　　金：国家自然基金青年科学基金项目(61101238);全军军事类研究生资助课题~~

摘　　要：目前,对无人机飞行声信号的分析主要是基于传统语音信号处理的手段,并未进行深入分析。文中针对无人机飞行声信号,结合无人机的气动特点深入研究分析得出无人机声信号的特征,分析比较傅里叶变换(FFT)、梅尔倒谱系数(MFCC)和基因周期3种特征提取算法并提取特征,应用支持向量机(SVM)分类算法,进行机型识别分类。实测与实验结果表明,FFT与MCC识别率相近,FFT运算复杂度低,基因周期不太适合单独进行特征识别,因此得出FFT适合作为无人机声特征提取方法。At present,the analysis of UAV flight acoustic signals is mainly based on traditional speech signal processing methods,but is not deep enough.In this paper,the UAV flight acoustic signal is studied and deeply analyzed in combination with the aerodynamic features of UAV to obtain the features of UAV acoustic signal.Three feature extraction algorithms of FFT(fast Fourier transform),MFCC(Mel frequency cepstral coefficient),and gene cycle are analyzed and compared,by which the features are extracted.The type of UAV is identified and classified by means of the SVM(support vector machine)classification algorithm.The measured results and experimental results show that the recognition rate of FFT and MCC is similar,the computa tional complexity of FFT is lower,and gene cycle is not suitable for feature recognition alone.Therefore,FFT is suitable for fea ture extraction of UAV acoustic signal.

关 键 词：无人机声信号 声信号特征提取 机型分类 算法分析 特征识别 实验分析 

分 类 号：TN912.3[电子电信—通信与信息系统] 34[电子电信—信息与通信工程]