基于CNN-LSTM的超宽带NLOS/LOS分类算法研究  

Research on Ultra-wideband NLOS/LOS Classification Algorithm Based on CNN-LSTM

作　　者：刘卿卿[1,2] 徐帅 王思语 吴南 刘明江 LIU Qingqing;XU Shuai;WANG Siyu;WU Nan;LIU Mingjiang(School of Automation,Nanjing University of Information Science&Technology,Nanjing 210044;Jiangsu Collaborative Innovation Center of Atmospheric Environment and Equipment Technology(CICAEET),Nanjing University of Information Science&Technology,Nanjing 210044)

机构地区：[1]南京信息工程大学自动化学院,南京210044 [2]南京信息工程大学江苏省大气环境与装备技术协同创新中心,南京210044

出　　处：《计算机与数字工程》2024年第8期2252-2258,共7页Computer & Digital Engineering

基　　金：国家自然科学基金项目(编号:61701244)资助。

摘　　要：超宽带(ultra-wideband,UWB)室内定位系统中,非视距(non-line-of-sight,NLOS)信号的识别与分类是影响定位精度的关键技术。为解决用于NLOS识别的信道冲激响应(Channel Impulse Response,CIR)中存在的噪声干扰问题以及传统NLOS识别方法多场景下阈值选取困难等问题,论文采用一种用于CIR数据去噪的可逆变换方法,并使用卷积神经网络(Convolutional Neural Network,CNN)结合长短期记忆网络(long short-term memory network,LSTM)的结构识别NLOS信号。采用开源数据集进行训练与验证,结果显示,CIR去噪后的CNN-LSTM分类精度最高可达90.62%。表明该方法可以有效提高NLOS信号分类精度。In ultra-wideband(UWB)indoor positioning systems,the identification and classification of non-line-of-sight(NLOS)signals is a key technology that affects positioning accuracy.In order to solve the problem of noise interference in the channel impulse response(CIR)used for NLOS identification and the difficulty of threshold selection in multiple scenarios of traditional NLOS identification methods,this paper adopts a reversible transform for CIR data denoising method,and uses convolutional neural network(CNN)combined with long short-term memory network(LSTM)structure to identify NLOS signals.The open source data set is used for training and verification.The results show that the classification accuracy of CNN-LSTM after CIR denoising can reach up to 90.62%,indicating that this method can effectively improve the classification accuracy of NLOS signals.

关 键 词：超宽带 非视距识别 去噪声 神经网络 

分 类 号：TN98[电子电信—信息与通信工程]