부산대학교 도서관

In the realm of underwater acoustic communication, where acoustic channels are characterized by Doppler spread, low SNR, and a scarcity of channel data, we present an Orthogonal Frequency Division Multiplexing (OFDM) communication scheme employing a Fully Connected Neural Network (FC-NN) to estimate the channel. This approach offers robust performance. Our FC-NN channel estimator is trained on a watermark channel and adapts to the dynamic underwater environment. Numerical results demonstrate low Bit Error Rate (BER) against traditional methods. We emphasize the FC-NN model’s resilience and its ability to maintain acceptable BER. This work significantly contributes to the reliability of underwater communication systems and holding promise for practical underwater communication applications.