结合时域分析和改进双谱的通信信号特征提取算法

发布时间：2018-06-19 04:52

本文选题：积分路径个数 + 改进矩形积分双谱　；参考：《信号处理》2017年06期

【摘要】：传统的矩形积分双谱特征存在以下不足:第一是在以往的研究中没有讨论过积分路径个数对识别率的影响;第二是在矩形积分双谱算法中存在着部分积分路径对识别效果贡献不足、甚至带来负作用的缺点。为克服这些问题、解决辐射源信号个体识别问题,本文提出了一种基于改进双谱和时域分析相结合的通信信号个体识别方法,首先通过实验得到了积分路径和识别率的性能曲线,选定最佳积分路径个数;其后重新定义Bhattacharyya距离,剔除掉对识别效果贡献不足、具有负作用的积分路径,得到具有最大比重的双谱区间;最后结合信号的时域特征并利用支持向量机分类器进行个体识别。本文分别通过仿真信号和实际信号对所提出算法做出了验证,实验结果表明,该方法能够较好解决辐射源信号的个体识别问题,平均正确识别率高于95%。
[Abstract]:The traditional bispectral features of rectangular integrals have the following disadvantages: first, the influence of the number of integral paths on the recognition rate has not been discussed in previous studies; The second is that the partial integral path has the disadvantage of insufficient contribution to the recognition effect and even negative effect in the rectangular integral bispectral algorithm. In order to overcome these problems and solve the problem of individual identification of emitter signals, this paper presents a method of individual recognition of communication signals based on the combination of improved bispectrum and time domain analysis. Firstly, the performance curves of the integral path and the recognition rate are obtained through experiments, and the optimal number of integral paths is selected. Then, the Bhattacharyya distance is redefined to eliminate the integral path which has negative effect and insufficient contribution to the recognition effect. The bispectral interval with the largest specific gravity is obtained. Finally, the time domain feature of the signal is combined and the support vector machine classifier is used to identify the individual. The experimental results show that this method can solve the individual recognition problem of emitter signal and the average correct recognition rate is higher than 95%.
【作者单位】：解放军信息工程大学信息系统工程学院;
【分类号】：TN911.6

【相似文献】