Abstract: Based on time-frequency analysis of the transmitters, a new transmitter individual identification method based on feature fusion is proposed. Firstly, the signal carrier frequency and transient amplitude characteristic are extracted, and then the fitting coefficients regarded as transient fingerprint characteristic are acquired by using segmented least squares curve fitting of the transient amplitude after resampling and cubic spline interpolation. Finally, the signal carrier frequency and the transient fingerprint characteristic are fused as the transmitter’s fingerprint feature vector as well as the fusion coefficients are optimized by genetic algorithms. Using probabilistic neural network classifier, the experiments are carried out based on 16 interphones. The experimental results show that the method is effectively, which is able to achieve available transmitter individual identification by mapping the signal time-frequency characteristics to the feature vectors, and the system recognition rate is above 90% with SNR of 20 dB.

Key words: transmitter identification, transient characteristics, feature fusion, probabilistic neural network

摘要： 在对辐射源信号进行时频分析的基础上，提出一种基于特征融合的通信辐射源个体识别方法。提取辐射源信号载频特征和瞬态幅值特征，对重采样的瞬态幅值做三次样条插值，采用最小二乘法分段对插值后的瞬态幅值进行曲线拟合，获取拟合系数作为瞬态指纹特征；最后与载频特征融合，采用遗传算法优化融合系数，融合后的特征作为辐射源指纹特征。识别分类器采用概率神经网络，对16部手持机进行识别实验。实验结果表明，该方法提取的特征能够反映通信辐射源个体的时频特性，可实现对辐射源个体的有效识别，在信噪比为20 dB时，系统识别率优于90%。

关键词: 辐射源识别, 瞬态特征, 特征融合, 概率神经网络