无线网络入侵干扰信号特征分离方法  

Feature Separation Method of Wireless Network Intrusion Interference Signal

作　　者：李闯 LI Chuang(China National Petroleum Power Co.,Ltd.,Daqing Heilongjiang 163000,China)

机构地区：[1]中国石油集团电能有限公司,黑龙江大庆163000

出　　处：《通信技术》2021年第10期2413-2417,共5页Communications Technology

摘　　要：由于对信号特征参数的提取目标缺乏针对性会导致特征分离精度较低,为此,提出了一种无线网络入侵干扰信号特征分离方法,建立了以短时傅里叶变换(Short-Time F ourier Transform,STFT)为基础的多源基函数。该方法将高频突发分量和长周期内的准平稳分量作为观测指标,在网络环境内,对信号断续和偏离量化处理,以此为基础对干扰信号的数学逻辑进行标识,提取标识信号的时域矩偏度、时域矩峰度、包络起伏度特征参数,并利用深度卷积神经网络实现提取特征的分离。实验结果表明,所提方法对12种常见干扰信号的特征分离精度可以达到95%以上,明显优于传统方法。Due to the lack of pertinence to the extraction target of signal feature parameters,the accuracy of feature separation will be low.Therefore,a feature separation method for wireless network intrusion interference signals is proposed,and a multi-source basis function based on short-time Fourier transform(STFT)is established.In this method,the high-frequency burst component and the quasi stationary component in the long period are taken as the observation indexes,and the signal discontinuity and deviation are quantified in the network environment.Based on this,the mathematical logic of the interference signal is identified,and the characteristic parameters of time-domain moment skewness,time-domain moment peak and envelope fluctuation of the identification signal are extracted,and the deep convolutional neural network is used to realize the separation of the extracted features.Experimental results indicate that the feature separation accuracy of the proposed method for 12 common interference signals can reach more than 95%,which is obviously better than traditional methods.

关 键 词：特征参数 特征分离 STFT 无线网络入侵 数学逻辑 时域矩峰度 

分 类 号：TN957[电子电信—信号与信息处理]