A Unified Convolutional Neural Network Classifier Aided Intelligent Channel Decoder for Coexistent Heterogeneous Networks

In coexistent heterogeneous wireless networks, receivers have to process the signaling and the data following different specifications. With the aim to automatically and intelligently identify the received signal type and then recover the data, in this paper, we propose a unified intelligent channel decoder serially concatenated by a convolutional-neural-network-based classifier and a deep learning (DL)-aided decoder. The classifier mainly consists of the convolutional layer, the batch normalization layer, and the max-pooling layer, while the DL decoder is constituted by four full connection layers. At the training stage, the unified decoder is trained to learn the distinct characteristics of encoded codewords following different specifications. Then, at the deployment stage, the classifier will extract the distinct structural features and identify the coding pattern. Thus, the DL decoder could choose an appropriate set of neural network parameters for information recovery. Simulation results demonstrate that our proposed intelligent decoder achieves better reliability performances than benchmark schemes over both additive white Gaussian noise and Rayleigh fading channels.