基于循环谱的去模糊调制识别算法  

Deblurring modulation recognition algorithm based on cyclic spectrum

作　　者：陈杰豪 王江[2] CHEN Jiehao;WANG Jiang(University of Chinese Academy of Sciences,Beijing 100049,China;Shanghai Institute of Microsystem and Information Technology,Chinese Academy of Sciences,Shanghai 200050,China)

机构地区：[1]中国科学院大学,北京100049 [2]中国科学院上海微系统与信息技术研究所,上海200050

出　　处：《现代电子技术》2024年第15期47-52,共6页Modern Electronics Technique

基　　金：国家重点研发计划项目:物联网智能传感器系统集成与应用示范(2021YFB3202105);中科院创新基金(CXJJ-23S037)。

摘　　要：针对当前调制识别算法在低信噪比下识别率低和循环谱应用中由于部分信号谱图相似而性能下降的问题,提出基于二维循环谱灰度图的去模糊调制识别算法。针对谱图特征相似的信号构建了二维循环谱灰度图模板库,通过与模板库进行谱图匹配将信号分流成两部分:谱图相似信号和谱图可区分信号。在信号分流基础上提出差异化识别方法,针对谱图可区分信号,通过构建基于二维循环谱灰度图的卷积神经网络(CSG-Net)完成识别;针对谱图相似信号,借助多通道学习深度神经网络(MCLDNN)完成识别。实验结果表明,提出的算法综合了不同网络的优势,提升了网络整体的识别性能,在-10 dB信噪比时依然有接近70%的识别率。In view of the low recognition rate of current modulation recognition algorithms at low signal-to-noise ratio(SNR)and its performance degradation of cyclic spectrum applications due to the spectrum similarity of some signals,a deblurring modulation recognition algorithm based on 2D cyclic spectrum grayscale is proposed.A 2D cyclic spectral grayscale template library is constructed for the signals with similar spectral features.And then,the input signals are streamed into two parts by spectral matching with the template library,including spectrum-similar signals and spectrum-distinguishable signals.A differentiated recognition method is proposed on the basis of signal streaming.For the spectrum-distinguishable signals,a convolutional neural network(CNN)based on 2D cyclic spectrum grayscale(CSG-Net)is constructed to complete the recognition.For the spectrum-similar signals,the recognition is completed with the help of multi-channel learning deep neural network(MCLDNN).The experimental results show that the advantages of different networks are integrated in the proposed algorithm,which improves the overall recognition performance of the network with a recognition rate close to 70%when the SNR is-10 dB.

关 键 词：循环谱 深度学习 卷积神经网络 自动调制识别 低信噪比 去模糊调制 

分 类 号：TN911.7-34[电子电信—通信与信息系统]