Artificial neural network estimation of data and channel characteristics in free-space ultraviolet communications

In this paper, we develop an artificial neural network (ANN)-based algorithm for signal classification, i.e., data demodulation, and the estimation of channel characteristics, such as channel DC gain and turbulence distribution parameters, in free-space ultraviolet (UV) communication systems. In this scheme, the ANN-based receiver adaptively tracks the UV channel variation and is directly trained using the data generated from UV channel models. To evaluate the performance of the proposed algorithm across all turbulence regimes, log-normal, Gamma-Gamma, and negative exponentially distributed UV turbulence channels are considered. We demonstrate that the proposed algorithm is robust to perform accurate and reliable signal classification and UV channel estimation without knowing underlying UV channel models or channel state information (CSI). In addition, time complexity analysis is presented. It is shown that, even under the strong turbulence regime, the accuracy of the proposed algorithm is found to be higher than 97% in terms of correct classification. It is also demonstrated that the classification error rate performance of the proposed ANN-based detector is superior to that of a maximum-likelihood (ML)-based detector with perfect CSI.