Two-Stage Channel Estimation Using Convolutional Neural Networks for IRS-Assisted mmWave Systems

Intelligent reflecting surface (IRS) is expected to be an essential component of next-generation wireless communication networks due to its potential to provide similar or higher array gain with lower hardware cost and energy consumption compared with massive multiple-input-multiple-output (MIMO) technology. However, channel estimation in IRS-assisted communication systems is more challenging than that in conventional systems. In this article, we propose a two-stage channel estimation approach using deep learning in millimeter-wave (mmWave) communication systems with a hybrid passive/active IRS structure. In the first stage, the sparsity of the mm Wave massive MIMO channel in the angular domain is exploited to estimate the amplitude of the sparse channel through a convolutional neural network. In this way, the indices of the nonzero entries of the sparse channel can be simultaneously obtained. In the second stage, the channel is reconstructed by solving a least squares problem with the acquired indices. The simulation results show that the proposed channel estimation scheme could achieve better performance with manageable complexity over the existing solutions.