Dual-Attention Based 3D Channel Estimation

Xiangzhao Qin; Sha Hu

cs.LGApr 2, 2026

Dual-Attention Based 3D Channel Estimation

Authors: Xiangzhao Qin, Sha Hu

Organizations: Lund Research Center, Huawei Technologies Sweden AB, Sweden.

Abstract

For multi-input and multi-output (MIMO) channels, the optimal channel estimation (CE) based on linear minimum mean square error (LMMSE) requires three-dimensional (3D) filtering. However, the complexity is often prohibitive due to large matrix dimensions. Suboptimal estimators approximate 3DCE by decomposing it into time, frequency, and spatial domains, while yields noticeable performance degradation under correlated MIMO channels. On the other hand, recent advances in deep learning (DL) can explore channel correlations in all domains via attention mechanisms. Building on this capability, we propose a dual attention mechanism based 3DCE network (3DCENet) that can achieve accurate estimates.

Figures & tables

Fig. 1: An AI receiver design comprised of AI based CE, MIMO detection, and log-likelihood ratios (LLRs) computations [ 1 , 2 ] .

Fig. 2: DMRS configurations for $N_{t}\!=\!4$ , $N_{f}\!=\!12$ , $N_{s}\!=\!14$ , and $N_{p}\!=\!3$ .

Fig. 3: The proposed 3DCE design that comprises a two-stage process of SA followed by TFA modules.

Module	Number of Weights	MACs per Batch
SA	$\begin{aligned} &(2N_{s}N_{c})d+2dd_{\mathrm{ff}}\\ &+d_{\mathrm{ff}}+2d+99\end{aligned}$	$\begin{aligned} &2N_{r}N_{t}(dN_{s}N_{c}+dd_{\mathrm{ff}}+49)\end{aligned}$
TFA	$\begin{aligned} &4d^{2}+5d+2dd_{\mathrm{ff}}+d_{\mathrm{ff}}\end{aligned}$	$\begin{aligned} &2L(2d^{2}+Ld+dd_{\mathrm{ff}})\end{aligned}$

TABLE I: Number of Parameter and MAC operations of the 3DCENet.

Fig. 4: CE-MSEs with different noise power allocations under correlated $4\!\times\!4$ MIMO and ETU100Hz channel.

Fig. 5: CE-MSEs with the proposed 3DCENet and other methods.

Fig. 6: BLER of $2\!\times\!2$ MIMO and 16QAM modulation.

Explore similar work

Improving Channel Estimation via Multimodal Diffusion Models with Flow Matching
Mar 13, 2026Xiaotian Fan, Xingyu Zhou, Le Liang +2Channel Estimation Multimodal Diffusion Models
School of Information Science and Engineering, Southeast University, Nanjing 210096, China · Purple Mountain Laboratories, Nanjing 211111, China
Multi-Block Attention for Efficient Channel Estimation in IRS-Assisted mmWave MIMO
May 14, 2026Mehrdad Momen-Tayefeh, Mehrshad Momen-Tayefeh, Maryam SabbaghianWireless Communications Channel Estimation
School of Electrical and Computer Engineering, University of Tehran, Tehran, Iran · Department of Computer Engineering, Sharif University of Technology, Tehran, Iran
Channel Estimation via Successive Denoising in MIMO OFDM Systems: A Reinforcement Learning Approach
Jan 25, 2021Myeung Suk Oh, Seyyedali Hosseinalipour, Taejoon Kim +2Channel Estimation
Electrical and Computer Engineering, Purdue University, West Lafayette, IN, USA · Electrical Engineering and Computer Science, University of Kansas, Lawrence, KS, USA

Mar 13, 2026J/K move · Enter open · S save

Improving Channel Estimation via Multimodal Diffusion Models with Flow Matching

Xiaotian Fan, Xingyu Zhou, Le Liang, Xiao Li+1

School of Information Science and Engineering, Southeast University, Nanjing 210096, China · Purple Mountain Laboratories, Nanjing 211111, China

Channel Estimation Multimodal Diffusion Models Flow-Based Generative Modeling

Open paper AskPDF arXiv

Deep generative models offer a powerful alternative to conventional channel estimation by learning complex channel distributions. By integrating the rich environmental information available in modern sensing-aided networks, this paper proposes MultiCE-Flow, a multimodal channel estimation framework based on flow matching and diffusion transformer (DiT). We design a specialized multimodal perception module that fuses LiDAR, camera, and location data into a semantic condition, while treating sparse pilots as a structural condition. These conditions guide a DiT backbone to reconstruct high-fidelity channels. Unlike standard diffusion models, we employ flow matching to learn a linear trajectory from noise to data, enabling efficient one-step sampling. By leveraging environmental semantics, our method mitigates the ill-posed nature of estimation with sparse pilots. Extensive experiments demonstrate that MultiCE-Flow consistently outperforms traditional baselines and existing generative models. Notably, it exhibits superior robustness to out-of-distribution scenarios and varying pilot densities, making it suitable for environment-aware communication systems.

Abstract

Explore similar work

Improving Channel Estimation via Multimodal Diffusion Models with Flow Matching

Multi-Block Attention for Efficient Channel Estimation in IRS-Assisted mmWave MIMO

Channel Estimation via Successive Denoising in MIMO OFDM Systems: A Reinforcement Learning Approach