Convolutional Neural Network Based Coherent Acoustic Signal Receiver Design for Underwater Acoustic Sensor Networks (UASNs)

In Underwater Acoustic Sensor Networks communication systems, synchronization is one of the most important problems. The need for synchronization is especially intensified when there is strong channel distortion. The Doppler effect can shift the carrier in the frequency domain and scale the signal in time domain. Like the multipath propagation channels poses problems to the conventional synchronization methods. The application is focused on the underwater acoustic communications, where the time varying multipath fading dominates the channel characteristics. The conventional synchronization techniques are generally derived based on maximum likelihood principle. In this project, a new entropy based synchronization technique is explored. Synchronization is achieved by minimizing the entropy estimated from the use of eye diagram and the constellation diagram. Key implementation details are addressed towards the realization of entropy based synchronization algorithms. In addition, the performance is evaluated in controlled conditions and that is shown that entropy minimization has great potential and offers certain advantages for synchronization in wireless communication, particularly for pulse shaping filters with small excess bandwidth, as well as in multipath fading channels. In final latest deep learning technique is applied to synchronize the Base band signal. A neural network based coherent receiver is designed and tested. Unlike the conventional receiver which consists of a series of function blocks, the neural network based (CNN) receiver does not explicitly implement in any function blocks. Its function is trained from end to end to achieve over all optimization. As such the new receiver structure has the potential to outperform the conventional receivers in nonlinear or nonparametric propagation channels.