Acquisition and feedback of accurate downlink (DL) channel state information (CSI) bring up high overhead. So a convolutional neural network (CNN) based multi-input-multi-output (MIMO) channel prediction method is proposed in this paper. Based on the space correlation, we design the proposed prediction method that predicts the MIMO channel concurrently. The proposed CNN method extracts the feature of UL-CSI through convolutional layers and predicts DL-CSI by the feature through deconvolutional layers. Simulation results demonstrate that the proposed channel prediction method leads to higher accuracy than existing methods. The bit error rate (BER) in orthogonal frequency division multiplexing (OFDM) system of proposed method is 82.4% and 26.2% lower than linear minimum mean squared error (LMMSE) method in two different simulated datasets. And the proposed method is tested under different user moving speeds and different antenna scale.


    Zugriff

    Zugriff prüfen

    Verfügbarkeit in meiner Bibliothek prüfen

    Bestellung bei Subito €


    Exportieren, teilen und zitieren



    Titel :

    Deep Learning-Based Time-varying Channel Prediction for MIMO Systems


    Beteiligte:
    Zhang, Shiyu (Autor:in) / Zhang, Yuxiang (Autor:in) / Zhang, Zhen (Autor:in) / Zhang, Jianhua (Autor:in) / Xia, Liang (Autor:in) / Jiang, Tao (Autor:in)


    Erscheinungsdatum :

    2022-06-01


    Format / Umfang :

    4235198 byte





    Medientyp :

    Aufsatz (Konferenz)


    Format :

    Elektronische Ressource


    Sprache :

    Englisch