DnCNN-based Sparse Channel Estimation For Mm-Wave MIMO-OFDM System

As one of the key technologies of 5G,massive MIMO has high spectral efficiency and energy efficiency.However,large-scale antennas at the base station side and complex frequency selective fading channels make accurate channel state information acquisition challenging.In view of the large number of estimation parameters for massive MIMO,most of the traditional channel estimation techniques are only applicable to 4G small-scale MIMO systems(such as 8-antenna LTE-A systems),and cannot be extended to massive MIMO systems directly.Therefore,this paper explores the estimation techniques applicable to 5G massive MIMO sparse channels.Based on the sparsity of beam domain and frequency correlation of wideband millimeter-wave massive MIMO channel,this paper proposes sparse channel estimation for wideband millimeter wave massive MIMO-OFDM system and further improvement scheme,using the residual network of deep learning and the theory of noise preprocessing.At first,based on the theory of beam squint in broadband systems and the related sparse structure in beam domain,this paper models the wideband MIMOOFDM system in the wideband beam domain and clarifies the basic principles of the Dn CNN residual learning network.Finally,the Dn CNN-based sparse channel estimation scheme is proposed and verified by experiments.performance.Secondly,based on the previous research,this paper proposes the Dn CNN sparse channel estimation scheme combined with noise preprocessing,which will further improve the performance of proposed algorithm.By denoising the input signal of Dn CNN,the algorithm complexity is reduced and the channel estimation performance is improved.