| 大会名称 |
| 2021年 総合大会 |
| 大会コ－ド |
| 2021G |
| 開催年 |
|
2021
|
| 発行日 |
|
2021-02-23
|
| セッション番号 |
|
BS-7
|
| セッション名 |
|
AI technologies and their applications for future network systems and services
|
| 講演日 |
|
2021/3/9
|
| 講演場所(会議室等) |
|
Meeting 30
|
| 講演番号 |
|
BS-7-1
|
| タイトル |
|
Comparative Evaluations of Deep Learning-based Super-Resolutions in Channel Estimation
|
| 著者名 |
|
◎Daiki Maruyama, Kenji Kanai, Jiro Katto,
|
| キーワード |
|
channel estimation, 5g system, super-resolution, deep learning, physical layer
|
| 抄録 |
|
Recently, deep learning (DL)-based radio communication systems are widely studied. For instance, a channel estimation method using DL-based super-resolution (SR), such as ChannelNet, has been proposed. In this paper, in order to enhance this research efforts, we apply newer and deeper SR networks to channel estimation. In addition, through 5G physical downlink simulation, we evaluate the performances of several estimation methods. From the evaluations, the results concluded that deeper SR networks are effective to estimate radio communication channel with higher accuracy and potentially reduce BER for radio communication.
|
| 本文pdf |
|
PDF download
|
