Deep Learning-Driven Joint Channel Estimation and Signal Detection for Underwater Wireless Optical Communication
A novel high-rate data service named underwater wireless optical communication (UWOC) method has been emerged recently. UWOC systems have to further consider the additional challenges in real-world situations where the transmission takes place underwater since the channel conditions are greatly complicated by the many effects of the water waves such as absorption, scattering, turbulence, which have different statistical traits in different kinds of water. These whimsical behaviors render both CE and SD nontrivial tasks to carry out reliable and fast transmissions over the channel. In this work, we propose a DL-based joint CC (class of channel), CE (channel estimation) and SD (signal detection) framework for UWOC systems. In contrarily most existing systems treat CE as a separate and often the first step in the process of retrieval of data. We show that CE can be integrated in SD using novel deep neural network (DNN) based design. They will be utilized for the CE process in the way that combinations of estimated weights (ECW) will be used adaptively in the presence of water of different types, potentially varying during the interaction which is most likely the case in the real world conditions. Simulation and experimental results reveal that the proposed system has great advantages of link performances in various UWOC channel environments.
