Deep Learning Enhanced Channel Estimation for the Internet of Vehicles OFDM Systems

The deep learning enhanced channel estimation is presented for the OFDM communication in the Internet of Vehicles (10V). Using the image enhancement and denoising, the channel responses at unknown positions are predicted under a specified Signal-to-Noise Ratio (SNR). Simulation experiments were conducted in predefined urban road scenes.Experimental results show that this method shows significant improvement compared to traditional methods in the minimum mean square error (MMSE) of channel estimation, with an average decrease of 0.03 in a series of experiments with different SNRs.