Bridging Neural Networks and Wireless Systems with MIMO-OFDM Semantic Communications
October 13, 2025 ยท View on GitHub
This repository contains the code for the paper
Bridging Neural Networks and Wireless Systems with MIMO-OFDM Semantic Communications.
Data and model checkpoints for the semantic server are available at Hugging Face.
If you find this repository or code useful, please consider citing our work:
@ARTICLE{yoo2025bridging,
author={Yoo, Hanju and Choi, Dongha and Kim, Yonghwi and Kim, Yoontae and Kim, Songkuk and Chae, Chan-Byoung and Heath, Robert W.},
journal={IEEE Wireless Communications},
title={Bridging Neural Networks And Wireless Systems with MIMO-OFDM Semantic Communications},
year={2025},
volume={32},
number={5},
pages={48-55},
month={Sept.}}
Directories
SERVER/: TCP-based server script for semantic encoding/decoding.
Requires TensorFlow with CUDA support. Model checkpoints must be downloaded from Hugging Face.CLIENT/: Baseband transmission scripts.- Sends input images to the server.
- Receives encoded symbols, modulates them with OFDM, and transmits via USRP.
- Performs reception, ZF equalization, symbol recovery, and decoding through the server.
Client PC must be connected to USRPs.
Requirements
- Server PC: TensorFlow, bpgenc
- Host PC: NumPy, SciPy, Ettus UHD driver, USRP X310 connection
- We use Jupyter Notebook extensions in VS Code to run
server.py/client.pywith an interactive GUI.
Run
- Launch
SERVER/ofdm_docker_server.pyon the server PC. - Configure
CLIENT/tcp_configs.pywith the server and USRP IP/port. - Run:
CLIENT/ofdm_semantic_client.py(semantic transmission)CLIENT/ofdm_bpg_client.py(BPG-based transmission)
Model Training
See SERVER/run.sh for training reference.