Intelligent and Reliable Coded Bit Stream Recovery over Correlated Fast Fading Channels

In this paper, we propose an intelligent coded-bits stream recovery scheme for improving reliability performance. In this design, a residual learning aided convolutional neural network (ResCNN) architecture with BCEwithLogits loss function is proposed to achieve the intelligent distortion compensation for the received data. In addition, at the online deployment stage, the ResCNN will calculate and output the float estimates of coded-bit streams, which act as the soft information to be regarded as the initial log likelihood ratio (LLR) for the subsequent decoding algorithm. Thus no channel state information (CSI) is required, then pilots used for the channel condition estimation can be removed to improve both the efficiency and the practicality. Subsequently, we analyze the computational complex and provide simulation results, which demonstrate that the proposed system can effectively improve the reliability performance of the receivers with no need of the CSI estimates to achieve higher practicality.