Frame Synchronization In Underwater Acoustic Communication Based on Convolutional Neural Network

In this article, frame synchronization based on convolutional neural network in the underwater acoustic channel is investigated and simulated. This study considers underwater acoustic channel problems including noise, Doppler phenomenon and multipath. Network training for frame synchronization shows a reduction in frame synchronization errors. Also, the use of auto-encoder in the input of the neural network of the synchronizer and using the code obtained from the auto-encoder reduces the learnable parameters of the neural network of the synchronizer, reduces the error and increases the resistance against the Doppler phenomenon.