水下冲激噪声环境下基于多特征融合的信号调制方式识别  

Classification of Signal Modulation Types Based on Multi-features Fusion in Impulse Noise Underwater

作　　者：赵自璐 王世练[1] 张炜[1] 谢阳[1] 

机构地区：[1]国防科学技术大学电子科学与工程学院,湖南长沙410073

出　　处：《厦门大学学报（自然科学版）》2017年第3期416-422,共7页Journal of Xiamen University：Natural Science

摘　　要：通信信号调制方式的识别在水下通信系统中发挥着重要作用,但目前在传统理论基础上建立起的高斯白噪声环境下的识别方法在水下冲激噪声背景下的识别仍存在困难.针对这一问题提出了水下冲激噪声环境下多特征融合的调制方式识别方法.利用Alpha稳定分布建立水下冲激噪声的模型,提出了基于指数函数的非线性变换方法消除部分冲激噪声的信号预处理方法;对预处理后的信号提取频域的盒维数特征、信号包络的样本熵特征以及Stockwell变换(S变换)域能量熵特征,构成多特征向量进行融合识别.对3个种类和数量不同的调制信号集进行仿真实验,结果表明,多特征融合识别的方法在水下冲激噪声环境下较单一特征识别性能更好,对Alpha稳定分布的特征指数在1~2之间时,该方法具有稳定性.同时通过对比仿真发现,非线性变换预处理显著地提高了算法性能,且多特征融合的调制识别方法的性能明显优于单特征方法,可识别的信号种类更多.Classification of signal modulation types plays an important role in underwater communication systems. However, presentclassical classification methods under white Gaussian noises exhibit poor performance under underwater impulse noises.This studyproposes a method of modulation classification based on multi-features fusion under underwater impulse noises.First,we apply thenonlinear transformation of the exponential function to eliminating part of impulse noises which are modeled by Alpha stable distri-bution model.Next, extract the signals~ frequency domain features of box dimension, envelopes features of sample entropy and Stock-well transform domain features of energy entropy,and then construct multi-features vectors, which are given to support vector ma-chine （SVM） for fusion and classification.Different signal sets are considered in MATLAB simulating experiments, which providedifferent number of signal schemes.Results show that the proposed method yields good performance and robustness under impulsenoise.In comparison with the simulation, it is verified that the pretreatment of nonlinear transformation significantly improves theperformance, and that the multi-features method offers better performance than those single ones do, at the same time increases the i-dentifiable signal types.

关 键 词：冲激噪声 调制识别 ALPHA稳定分布 多特征 

分 类 号：TN929.3[电子电信—通信与信息系统]