基于小波熵的辐射源指纹特征提取方法  

Fingerprint Feature Extraction Method for Emitters Based on Wavelet Entropy

作　　者：徐玉龙[1] 王金明[1] 徐志军[1] 陈志伟[1] 周坤[1] 

机构地区：[1]解放军理工大学通信工程学院,南京210007

出　　处：《数据采集与处理》2014年第4期631-635,共5页Journal of Data Acquisition and Processing

摘　　要：在对辐射源信号进行小波分析的基础上,提出一种基于小波熵的辐射源指纹特征提取方法。首先计算辐射源信号的功率谱,对功率谱进行连续小波变换,提取不同尺度下小波系数的熵特征作为辐射源信号指纹特征。识别分类器采用概率神经网络,对20部手持机进行识别实验,并与传统矩形积分双谱进行对比。实验结果表明,该方法能够把辐射源信号的时频特性信息通过小波系数的熵特征映射到特征向量中,从而实现对辐射源个体的有效识别,而且该特征参数对噪声干扰不敏感,在信噪比为20dB时,系统识别率达到95%以上,在信噪比为5dB时系统识别率仍优于80%,验证了所提方法的有效性。Based on wavelet analysis of the emitters,a new fingerprint feature extraction method for emitter identification based on wavelet entropy is proposed.Firstly,the signal power spectra are calculated.Secondly,the wavelet coefficients are extracted by continue wavelet transform.Finally,the wavelet entropy is extracted as a feature vector.Using neural network classifier,the comparative experiments with traditional square integral bispectrum are carried out based on twenty interphones.The experimental results show that the method can achieve individual classification by transferring the signal time-frequency characteristics to the feature vectors through the entropy of the wavelet coefficients.Besides,the proposed method is insensitive to noise,and the system recognition rate is above 95％ and more than 80％ with SNRs of 20 dB and 5 dB,respectively.

关 键 词：辐射源识别 小波熵 指纹特征 

分 类 号：TN911.72[电子电信—通信与信息系统]