信道状态信息(Channel State Information,CSI)反馈是大规模多输入多输出(Multiple-Input Multiple-Output,MIMO)系统的一个关键问题.大规模MIMO系统中基站天线数量巨大,CSI反馈出现了反馈开销大、反馈精度低等问题.为了降低反馈开销,提高反馈精度,采用深度学习方法,提出了一种基于特征融合的CSI反馈网络(Feature Fusion Net,FFNet).利用基于注意力机制的特征融合在编码器中融合不同尺度的CSI特征,并在解码器中使用多通道多分辨率卷积网络以及通道重排,从而高精度地重建压缩后的CSI.仿真结果表明,与几种经典的深度学习CSI反馈方法相比,在室内和室外信道条件下,均具有更高的反馈精度.
CSI feedback for large-scale MIMO systems based on feature fusion
Channel state information(CSI)feedback is a key issue in large-scale multiple-input multiple-output(MIMO)systems.The number of base station antennas in large-scale MIMO systems is huge,and the CSI feedback holds problems such as large feedback overhead and low feedback accuracy.In regard of these,a feature fusion-based CSI feedback network,FFNet,is proposed based on a deep learning approach.The CSI features are fused at different scales in the encoder,while an attention-based mechanism of feature fusion,a multi-channel multi-resolution convolutional network,and the channel rearrangement are deployed in the decoder.Thus,the compressed CSI is reconstructed with high accuracy.Simulation results show that the feedback accuracy is higher in both indoor and outdoor channel conditions,compared to several classical deep learning CSI feedback methods.
large-scale multiple-input multiple-output(MIMO)channel state information(CSI)deep learningconvolutional neural networkfeature fusion
国家科技期刊平台
NSTL
万方数据
