关于我About Me
张泽中，2017年6月本科毕业于南方科技大学，2021年9月获得香港大学电气与电子工程系博士学位，2021年11月至2023年12月于香港中文大学（深圳）未来智联网络研究院课题组任博士后研究员，2024年1月加入香港中文大学（深圳）理工学院任研究助理教授。
Zezhong Zhang received his B.Eng. degree from Southern University of Science and Technology in June 2017, and his Ph.D. degree from the Department of Electrical and Electronic Engineering at The University of Hong Kong in September 2021. From November 2021 to December 2023, he was a Postdoctoral Research Fellow at the Future Networks of Intelligence Institute, CUHK Shenzhen. In January 2024, he joined the School of Science and Engineering at CUHK Shenzhen as a Research Assistant Professor.
我的研究兴趣包括无线电地图感知、边缘学习、生成式人工智能以及语义通信等6G技术，当前主要研究方向为基于生成式人工智能技术的低空无线电地图&信道知识地图构建，该课题由深圳市未来智联网络研究院支持，同时本人参与并主要推动香港中文大学（深圳）牵头的广东省电磁频谱科学数据中心项目。如有意向参与课题组科研，欢迎联系：zhangzezhong@cuhk.edu.cn
My research interests include radio map sensing, edge learning, generative AI, and semantic communication and other 6G technologies. Currently, my main research focuses on low-altitude radio map and channel knowledge map construction based on generative AI technologies, supported by the Future Networks of Intelligence Institute, CUHK Shenzhen. I am also actively involved in and driving the Guangdong Provincial Electromagnetic Spectrum Science Data Center project led by CUHK Shenzhen. For collaboration inquiries, please contact: zhangzezhong@cuhk.edu.cn
研究方向Research Interests
点击研究方向查看相关论文
Click on a research area to view related papers
🗺️ 无线电地图感知Radio Map Sensing
3D无线电地图估计、生成扩散模型、无线环境感知
3D Radio Map Estimation, Generative Diffusion Models, Wireless Environment Sensing
4 篇论文4 Papers🤖 边缘学习Edge Learning
联邦学习、分布式机器学习、边缘智能
Federated Learning, Distributed Machine Learning, Edge Intelligence
3 篇论文3 Papers📡 B5G/6G通信B5G/6G Communication
通感一体化、大规模MIMO、智能反射面
ISAC, Massive MIMO, Intelligent Reflecting Surface
4 篇论文4 Papers🔗 语义通信Semantic Communication
语义通信、知识库增强通信、深度学习在无线通信中的应用
Semantic Communication, Knowledge Base Enhanced, Deep Learning for Wireless
3 篇论文3 Papers发表论文Publications
🗺️ 无线电地图感知Radio Map Sensing
无线电地图是描述无线信号强度在空间分布的数字地图，在6G通信中扮演关键角色。我们团队在3D无线电地图生成与估计领域开展了系统深入的研究工作：
• RadioGen3D：基于2D传播模型和实测数据，通过回归拟合的方法灵活高效地构建了高质量3D无线电地图数据，形成了大规模3D无线电地图合成数据集，适合用于3D AI模型的预训练，为解决3D地图数据稀缺问题提供了新思路。同时，基于Radio3DMix数据集，提出了一种基于GAN方法的3D UNet模型训练方法，训练后的3D UNet模型能够实现高精度三维无线电地图的实时构建，并已应用部署于真实校园场景
• RadioDiff-3D：首个3D×3D无线电地图数据集，结合生成扩散模型构建基准测试平台，为6G环境感知通信研究提供重要支撑
• 隐私敏感场景的分布式频谱感知：利用截断垂直联邦学习实现低延迟协同频谱感知，兼顾隐私保护与感知效率
• 生成式无线电地图构建：通过GAN实现快速准确的协同无线电地图估计
Radio map describes the spatial distribution of wireless signal strength and plays a key role in 6G communications. Our team has conducted systematic research on 3D radio map generation and estimation:
• RadioGen3D: Based on 2D propagation models and measured data, efficiently constructs high-quality 3D radio map data through regression fitting, forming a large-scale 3D radio map synthetic dataset suitable for pretraining 3D AI models, providing a novel approach to solve 3D map data scarcity. Moreover, based on the Radio3DMix dataset, we propose a GAN-based 3D UNet training method. The trained 3D UNet model enables real-time construction of high-precision 3D radio maps, already deployed in real campus scenarios
• RadioDiff-3D: The first 3D×3D radio map dataset, combined with generative diffusion models to build a benchmark platform for 6G environment-aware communication research
• Privacy-Preserving Distributed Spectrum Sensing: Low-latency cooperative spectrum sensing via truncated vertical federated learning, balancing privacy protection and sensing efficiency
• Generative Radio Map Construction: Fast and accurate cooperative radio map estimation via GAN
🤖 边缘学习Edge Learning
边缘学习将机器学习推向网络边缘，在保护数据隐私的同时实现分布式智能。我们的研究重点包括：
• 联邦学习拓扑优化：通过优化设备间的通信拓扑结构，加速联邦学习收敛
• 卫星-地面网络联邦学习：通过专家驱动的模型分割实现星地网络下的联邦调参
• 空中计算：创新性地利用信道噪声作为计算资源，实现高效的空中PCA，将"噪声"变"加速器"
我们的方法显著降低了通信开销和延迟，同时保护了用户数据的隐私安全。
Edge learning pushes machine learning to the network edge, enabling distributed intelligence while preserving data privacy. Our research focuses on:
• Federated Learning Topology Optimization: Accelerating FL convergence by optimizing communication topology
• Satellite-Terrestrial Federated Learning: Expert-driven model splitting for federated tuning in satellite-terrestrial networks
• Over-the-Air Computation: Innovatively leveraging channel noise as computational resource for efficient over-the-air PCA, turning "noise" into "accelerator"
Our methods significantly reduce communication overhead and latency while preserving data privacy.
📡 B5G/6G通信B5G/6G Communication
面向下一代（B5G/6G）无线通信系统，我们在大规模MIMO、毫米波通信和智能反射面等关键技术方向展开深入研究：
• 大规模MIMO：研究D2D网络下的速率自适应、导频污染抑制等挑战，提出基于学习的优化方法
• 毫米波通信：利用毫米波表面反射实现多节点车辆定位，赋能自动驾驶
• 智能反射面：通过智能调控电磁波传播，提升系统覆盖和能效
这些技术将为6G网络的自适应、高效能通信奠定基础。
For next-generation (B5G/6G) wireless communication systems, we conduct in-depth research on key technologies including massive MIMO, millimeter-wave communications, and intelligent reflecting surfaces:
• Massive MIMO: Studying rate adaptation in D2D networks and pilot contamination suppression with learning-based optimization
• Millimeter-wave Communications: Multi-point vehicular positioning via mmWave surface reflection for autonomous driving
• Intelligent Reflecting Surface: Smart electromagnetic wave propagation control for enhanced coverage and energy efficiency
These technologies will lay the foundation for adaptive, high-efficiency 6G networks.
🔗 语义通信Semantic Communication
语义通信是6G的范式转变，从比特传输转向语义传递。我们的研究深入探讨：
• 语义通信的本质：我们发表在《通信与信息Networks》的综述论文被引用333次，系统性地阐述了语义通信如何在未来智能时代传递"意义"而非"比特"
• 知识库增强的语义通信：利用外部知识库辅助语义编码和解码，显著提升语义传输的准确性和效率
• 语义安全：在语义层面实现通信安全防护，兼顾传输效率与安全性
语义通信有望从根本上改变人机交互的方式，实现更加智能、高效的通信系统。
Semantic communication represents a paradigm shift in 6G, from bit transmission to meaning delivery. Our research explores:
• Nature of Semantic Communication: Our survey (333 citations) systematically explains how SC conveys "meaning" instead of "bits" in the era of machine intelligence
• Knowledge Base Enhanced SC: Using external knowledge bases to assist semantic encoding/decoding for improved accuracy and efficiency
• Semantic Security: Achieving communication security at the semantic level, balancing efficiency and security
Semantic communication promises to fundamentally transform human-machine interaction for smarter, more efficient systems.