Deep Residual Learning for Channel Estimation in Intelligent Reflecting Surface-Assisted Multi-User Communications

Channel estimation is one of the main tasks in realizing practical intelligent reflecting surface-assisted multi-user communication (IRS-MUC) systems. However, different from traditional systems, an IRS-MUC system generally involves a cascaded channel with sophisticated statistical distribution. In this case, optimal minimum mean square error (MMSE) estimator requires calculation multidimensional integration which intractable to be implemented practice. To further improve performance, paper, we model as denoising problem and adopt deep residual learning (DReL) approach implicitly learn noise for recovering coefficients noisy pilot-based observations. end, first develop versatile DReL-based framework where network (DRN)-based MMSE derived terms Bayesian philosophy. As realization developed DReL framework, convolutional neural (CNN)-based DRN (CDRN) then proposed CNN block equipped element-wise subtraction structure specifically designed exploit both spatial features matrices additive nature simultaneously. particular, explicit expression CDRN analyzed characterize its properties theoretically. Finally, simulation results demonstrate that performance method approaches requiring availability prior probability density function channel.