Data-Aided Receiver for Time-varying SIMO-OFDM Systems

Deep learning (DL) is now widely used in the physical layer of wireless communications. Through multi-layer neural networks, it extracts the implicit features from the training data and simulate the non-linear mapping between the input and the output. In this paper, we focus on uplink timevarying single-user orthogonal frequency division multiplexing (OFDM) systems and propose a novel data-aided receiver based on DL. Specifically, we adopt superimposed pilot to enhance the data transmission efficiency and avoid the large pilot overhead in time-varying systems. The proposed scheme in this work not only utilizes the powerful learning ability of DL and adopts datadriven modules such as residual block, but also considers the potential of data information and integrates the idea of iterative receiving structure in traditional digital receiver, effectively improving the training performance and achieving more reliable data transmission. Simulations verify that the data-aided DLbased approach proposed in this paper significantly outperforms the traditional non-data-aided receivers. Keywords-OFDM, time-varying channel, model-driven deep learning, data-aided