Channel Estimation by Using Very Deep Super-Resolution Technique for OFDM-Based Underwater Acoustic Communication Systems

This paper presents a channel estimation method that uses Very Deep Super Resolution (VDSR) technique for OFDM-based underwater acoustic communication systems. Our approach consists of two steps. In the first step, we gather measured data of the underwater acoustic channel (UAC) in a real underwater environment. This data is then used to train a neural network. In the second step, we verify the information of a real transmission using the OFDM technique. The received signal is demodulated using a conventional technique, whereby the channel is estimated using the Very Deep Super Resolution (VDSR) network. This is a novel technique proposed in this paper. Simulation and experimental results demonstrate that using the neural network in combination with the VDSR technique for channel estimation is more effective than traditional approaches such as the LS method, as shown by the lower bit error rate (BER).