Summary
International Conference on Emerging Technologies for Communications
2022
Session Number:O2
Session:
Number:O2-4
LOS/NLOS Classification for Downlink CDL Channel Using Supervised Learning
Jingyu Liu, Mondher BOUAZIZI, Tomoaki Ohtsuki,
pp.-
Publication Date:2022/11/29
Online ISSN:2188-5079
Summary:
Transfer learning-based method channel state information (CSI) feedback can achieve a good CSI reconstruction accuracy with small amount of computation time and cost. However, it needs to select a proper source channel model. To find out the proper source channel model, classifying line-of-sight (LOS) and none-line-of-sight (NLOS) is an important step. Most existing works for LOS/NLOS classification focus on the ultra wide band (UWB) systems. In this paper, we use supervised learning to classify the downlink clustered delay line (CDL) channels between LOS and NLOS with the help of selecting feature sets. We also classify the CDL channels as a multi-class: CDL-A, CDL-B, CDL-C, CDL-D, and CDL-E. The results show that the classification for LOS/NLOS scenarios reaches 95%. In the case of multi-class classification, the accuracy reaches around 67%.