基于高阶累量与神经网络覆盖算法的数字调制识别  

A Method of Modulation Classification Based on HOC and a Covering Algorithm for Constructing Neural Networks

作　　者：张金涛[1] 解明祥[1] 

机构地区：[1]解放军电子工程学院信息系,合肥230037

出　　处：《电子信息对抗技术》2006年第5期15-18,共4页Electronic Information Warfare Technology

摘　　要：提出了一种把归一化四阶累量作为分类特征参数,应用神经网络覆盖算法进行分类的调制识别算法。对MFSK信号的三种归一化四阶累量进行了理论推导,对几种不同调制方式的数字通信信号的三种特征参数随信噪比的变化进行了仿真,证明了高阶累量不受高斯噪声影响的理论结果。最后应用神经网络覆盖算法进行分类识别,得到了比较好的识别结果。A new method of modulation classification for digital communication signals, using normalized fourth-order cumulants as characteristics and a covering algorithm for constructing neural networks as a classifier, is proposed. Normalized fourth-order cumulants of MFSK is calculated. Characteristics parameters of several kinds of digital communication signals versus SNR is simulated, which testifies HOC is insensitive to AWGN. At last, a covering algorithm for constructing neural networks is used to classify different modulation signals, and the good result of classifications is obtained by computer simulation.

关 键 词：四阶累量 神经网络覆盖算法 数字通信信号 调制识别 

分 类 号：TN911.7[电子电信—通信与信息系统]