频道(广播)
计算机科学
算法
扩散
估计理论
控制理论(社会学)
噪音(视频)
数学
应用数学
理论(学习稳定性)
计算
稳健性(进化)
数学优化
估计
工作(物理)
约束(计算机辅助设计)
期限(时间)
作者
Junhao Li,Z. Y. Chang,Jiali Zeng,Yuehong Gao
出处
期刊:
[Institute of Electrical and Electronics Engineers]
日期:2025-12-12
卷期号:: 1477-1482
标识
DOI:10.1109/iccc68654.2025.11437826
摘要
Accurate channel estimation is important for the performance of modern wireless systems, especially in high-frequency and large-scale multiple-input multiple-output (MIMO) scenarios. Traditional model-based methods often fail in dynamic or complex environments due to limited adaptability. In this work, we propose a location-aware Conditional Diffusion Models (CDMs) for robust and accurate channel estimation and generation. By incorporating user position as conditional input, our model learns to reconstruct channel matrices from noisy observations through a forward noise injection and reverse denoising process. We further introduce a Feature-wise Linear Modulation (FiLM) based modulation mechanism to enhance spatial feature learning. Experiments on the DeepMIMO dataset across mixed scenario demonstrate that our method significantly outperforms conventional baselines, especially in low signal-tonoise ratio (SNR) environments. Our approach offers a datadriven alternative to conventional methods and provides new insights for channel-aware intelligent communication systems.
科研通智能强力驱动
Strongly Powered by AbleSci AI