Under the condition of low signal-noise-ratio, the individual features of communication transmitter in steady-state signal are covered very easily, and hard to be extracted and identified. In regard to this problem, considering the fact that oscillators used in different communication transmitters have unequal frequency stability, a new method based on fractal dimension and Support Vector Machine (SVM) was proposed. After oversampling IF ( Intermediate Frequency) signal, information dimension was extracted as features, and then SVM classifier was designed to realize the automatic identification of unknown samples. To some extent, the feature was robust under AWGN (Additive White Gaussian Noise). The computer simulation shows that the method has good performance on classifying five PSK signals with the same order and the frequency stability difference of 0.01 ppm under 3 Db, its accuracy being 95%.%在低信噪比条件下,稳态信号中的通信辐射源个体特征极易被掩盖,从而难以提取和识别.针对该问题,依据不同通信辐射源中振荡器个体的频率稳定度不相等这一事实,提出一种适用于多进制数字相位调制(MPSK)信号的基于分形维数的特征提取与分类方法.首先对中频信号进行过采样,然后提取信号瞬时相位的信息维数作为分类特征,最后利用支持向量杌(SVM)分类器实现样本属性的自动判别.该方法特征维数低、分类简单,对加性高斯白噪声具有一定的鲁棒性.计算机仿真实验结果表明,当信噪比为3dB时,对载波频率稳定度差异为0.01 ppm的5种同阶MPSK信号的平均分类准确率达到95％.
展开▼
