基于深度学习的噪声背景通信信号端点检测  

End-point detection for noise background communication signal based on deep learning

作　　者：程涛 姚万华 姚克 栗高尚 Cheng Tao;Yao Wanhua;Yao Ke;Li Gaoshang(The 28th Research Institute of China Electronics Technology Group Corporation,Nanjing 210028,China)

机构地区：[1]中国电子科技集团公司第二十八研究所,江苏南京210028

出　　处：《无线互联科技》2023年第2期13-19,共7页Wireless Internet Technology

摘　　要：在复杂的电磁和地理自然环境中,通信链路常常受到噪声的干扰。基于此,文章提出了一种基于深度学习的方法来解决噪声背景下的通信信号端点检测问题,该方法使用深层卷积神经网络(Deep Convolutional Neural Networks,DCNN)提取样本特征,用于描述信号活跃区域和背景噪声之间的差异,并获得样本特征图。同时,通过多尺度区域检测方法确定特征图中的通信信号的起止端点,并使用线性回归方法精修区域参数,使端点检测结果更接近真值。在实验验证方面,文章利用构建的仿真通信信号数据集进行训练和测试,实验结果表明,该方法能够在毫秒级延迟下准确地检测出淹没在噪声中的通信信号,且检测精度优于现有方法。In complex electromagnetic and geographical natural environments,communication links are often disturbed by noise.Based on this,this paper proposes a method based on deep learning to solve the problem of communication signal endpoint detection in the background of noise.This method uses deep convolutional neural network(DCNN)to extract sample features to describe signal activity.difference between the region and the background noise,and obtain sample feature maps.At the same time,the start and end endpoints of the communication signals in the feature map are determined by the multi-scale region detection method,and the region parameters are refined by using the linear regression method,so that the endpoint detection results are closer to the true value.In terms of experimental verification,the simulated communication signal data set constructed in this paper is used for training and testing.The experimental results show that the method can accurately detect communication signals submerged in noise with a delay of milliseconds,and the detection accuracy is better than that of existing methods.method.

关 键 词：端点检测 深度学习 卷积神经网络(CNN) 边框回归 

分 类 号：V221+.3[航空宇航科学与技术—飞行器设计] TB553[理学—物理]