基于注意力机制引导深度残差网络的RIS辅助通信信道估计

Channel Estimation for RIS-aided Communications Based on Attention Mechanism-guided Deep Residual Networks

用可重构智能表面(Reconfigurable Intelligent Surface, RIS)增强无线覆盖和信道容量是未来通信网络的候选方案之一。为估计RIS辅助的多用户(Multi User, MU)通信系统上行链路的信道状态信息,提出一种基于注意力机制的深度残差网络,构建了包含稀疏块、特征增强块、注意力引导块和重构块的网络结构,隐式地学习残差噪声,利用注意力机制加强对特定信道噪声特征的提取。仿真结果表明,该方法的估计精度略低于线性最小均方误差(Linear Minimum Mean Square Error, LMMSE)估计,在高信噪比时比常规深度残差去噪网络的估计精度更高。

Using Reconfigurable Intelligent Surface(RIS)to improve wireless coverage and channel capacity has been considered as one of the candidates for future wireless communications.In order to estimate the state information of uplink Multi User(MU)channel,an attention mechanism-based deep residual network is proposed.Its structure is constructed which is comprised of a sparse block,a feature enhancement block,an attention block and a reconstruction block.The network not only implicitly learns residual noise,but also uses attention mechanism to enhance the extraction of specific channel noise features.Simulation results show that the method’s accuracy is comparable to that of the ideal Linear Minimum Mean Square Error(LMMSE)estimation,whereas higher than that of a general deep residual denoising network at high signal to noise ratio.