| 摘要: |
| 针对超大规模多输入多输出（Extremely Large-scale Multiple-Input Multiple-Output，XL-MIMO）系统中因天线规模庞大而导致信道状态信息（Channel State Information，CSI）难以高效获取，以及实际系统中信道稀疏度难以获取的问题，提出了一种基于随机梯度追踪(Stochastic Gradient Pursuit，SGP)算法的混合场信道估计算法，在信道稀疏度未知场景下，以Dice系数准则替代传统内积准则来提高原子选择的正确性，从而有效实现混合场信道的估计。数值结果显示，与基于分段弱正交匹配追踪的信道估计算法相比，所提混合场信道估计算法在信道估计性能上提升了约2.2 dB，验证了该算法在信道稀疏度先验知识未知的情况下仍可显著提高信道估计精度与系统鲁棒性。 |
| 关键词: 超大规模MIMO 信道估计 信道状态信息 随机梯度追踪 |
| DOI：10.20079/j.issn.1001-893x.250514001 |
|
| 基金项目:重庆市自然科学基金创新发展联合基金（中国星网）（CSTB2023NSCQ-LZX0114） |
|
| Hybrid-field Channel Estimation Based on Dice-guided Stochastic Gradient Pursuit |
| GUO Xiaojin,LIU Yuke |
| (School of Communications and Information Engineering，Chongqing University of Posts and Telecommunications，Chongqing 400065，China) |
| Abstract: |
| To address the challenge of efficiently acquiring channel state information(CSI) in extremely large-scale multiple-input multiple-output(XL-MIMO) systems caused by the massive antenna array，as well as the difficulty of obtaining channel sparsity in practical deployments，a hybrid-field channel estimation algorithm based on the Stochastic Gradient Pursuit(SGP) method is proposed.In scenarios where the channel sparsity is unknown，the proposed algorithm replaces the conventional inner product criterion with a Dice similarity-based selection rule to enhance the accuracy of atom selection，thereby enabling effective hybrid-field channel estimation.Numerical results show that，compared with the channel estimation algorithm based on Segmented Weak Orthogonal Matching Pursuit(SWOMP)，the proposed method achieves an improvement of approximately 2.2 dB in estimation performance，demonstrating that it can significantly enhance estimation accuracy and system robustness even without prior knowledge of channel sparsity. |
| Key words: extremely large-scale MIMO channel estimation channel state information stochastic gradient pursuit |
《电讯技术》编辑部 