Design of Smoothed L0 UnfoldingNet Architecture for OFDM Channel Estimation

Recently, the sparse recovery (SR) techniques could be used to achieve better channel estimates with fewer pilot resources in the OFDM system. To improve the priori information of SR techniques, in this paper, the smoothed LO UnfoldingNet (SLO-UNet) architecture is designed. In the proposed architecture, two different learning methods are presented to enhance the recovery performance for channel estimation. The results show that both two proposed learning methods could achieve more robust performance with limited complexity increased compared to the traditional SR technique.