【作者】 李琳；

【导师】 裴昌幸； 龚晓峰；

【作者基本信息】 西安电子科技大学 ， 电子与通信工程， 2012， 硕士

【摘要】 无线电监测,指的是通过探测、搜索、截获无线电管理领域内的无线电信号,由此对该无线电信号进行分析、识别、监视并获取其技术参数、工作特征等技术信息的活动。而信号调制方式识别,就是在未知调制信息内容的情况下对给定接收通信信号的调制方式进行判断,并估计给出相应的调制参数。要想监视无线电信号,首先就是要能分析识别出调制信号的类型特征,因而信号的调制方式识别无线电监测处理中起着基础且十分重要的地位。研究已有各类识别技术与算法的基础上,本文主要做了以下工作：介绍了常见的12种调制信号,并给出各类调制信号的数学模型,通过信号仿真讨论了各类调制信号瞬时幅值,瞬时相位及瞬时频率的特征。结合无线电监测应用背景讨论了噪声对调制信号的影响。分析、比较了当前信号调制方式识别各类算法的优缺点,着重研究了小波变换对调制信号进行降噪预处理算法；小波变换对调制信号进行突变边界特征提取算法；零中心归一化瞬时幅度之谱密度最大特征提取算法以及核判别分析算法。给出综合利用上述算法,结合组合分类的思想对各类调制信号进行逐层提取,最终实现各类调制信号的完整分类。在仿真试验中,当信噪比为5dB时,识别率可以达到90%以上,基本能完成所给各类调制信号的自动调制识别。该结果表明了组合分类的策略在低信噪比情况下分类结果仍能达到满意的效果,证明了本文所采用综合算法的有效性。更多还原

【Abstract】 Radio signal monotoring refers to detect, search and receive modulated radio signals, then analysis, identify and monitor the feature parameters of signals. To the process of signal modulation, it means that judging and estimating the characteristic parameters of modulated signals without any priori informations. The first step of monotoring radio signals is analysising and estimating the features of modulated signals, therefore modulation recognition of siganals plays an important role in the field of radio signal monitoring.This paper studies the modulation recognition with a genaral strategy of combining advanced algorithms. First we introduce12kinds of common communication signals, and gives the features of instant amplititude, instant phase and instant frequency and so on. We also disscuss the impact of noises to the modulated signals. After analysising and comparing different signals modulation recognition algorithms, first we focus on studying the wavelet transform on denoising and the selection of signal’s abrupt boundary, and then analysis the advantages of the maximum spectral density feature method and the Kernel fisher discriminant algorithm. With a synthesize strategy of combining these algorithms, we can solves the distinguishability to all the signals step by step. In the simulation experiments parts, the results illustrates that the multi-class signals identification algortihm can achieves above90%precision under the case of low SNR by choosing appropriate wavelet function and threshhold value. It also proves that this classifier strategy has a high and steady success rate, especially to the case of low SNR(0dB～5dB) in the computer simulation experiments.更多还原