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.
Abstract
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 chan-nel 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.
维普
万方数据