基于分形特征提取的调制信号识别  

Modulation Classification Based on Fractal Features

作　　者：林俊[1] 熊刚[2] 王智学[1] 

机构地区：[1]解放军理工大学通信工程学院,南京210007 [2]南京理工大学引信抗干扰实验站,南京210094

出　　处：《弹箭与制导学报》2004年第S6期358-363,共6页Journal of Projectiles,Rockets,Missiles and Guidance

摘　　要：研究了非平稳分形随机噪声背景下的信号调制识别问题。提取了盒维数、信息维数和高阶分形特征-缝隙尺度变化率构成分类特征向量空间,这些特征包含了信号幅度、频率和相位的变化规律,集中了各种调制方式之间的差异信息,通过理论分析和仿真表明,当 SNR>5dB 时,(1)分形特征对分形随机噪声不敏感,能充分保证分类特征在决策空间存在一个较清晰的边界;(2)在相同的 SNR 环境下,分形特征在分形噪声环境下比在高斯白噪声环境下有更好的稳定性,对5种调制样式的识别率均大于92.5%。The modulation classifation in the non-stationary fraetal stochastic noise was studied.This paper extracted the box dimension,information dimension and the high-order fractal feature-aperture resolution slope as the feature vector, which contain the change law of amplitude,frequency and phase and diversity between different modulation classes.The theoretic analysis show when Snr>5dB:(1)the affection of fractal noise is so light that the classification feature retain a distinct boundary.(2)The fractal feature is more steady in the fractal noise than in the Gauss noise with the same Snr,and the all the ratio of reorganization is above 92.5%.

关 键 词：高阶分形特征 分形噪声 特征提取 调制识别 

分 类 号：TN76[电子电信—电路与系统]