Channel Coding and Interleaving Blind Recognition using Deep Learning

This paper proposes two deep learning models for recognizing the types of channel coding and interleaving in a blind scenario, where no information about the transmitted signal is available. For the success of military operations, it is crucial to intercept and analyze enemy communication signals, which requires the ability to blindly recognize the channel coding and interleaving techniques used by the adversary. The signals used in this study are random messages that have passed through channel coding, interleaving, and scrambling. Simulation results show that when the ratio of Os to 1s in the message is 50:50, the deep neural network (DNN) achieves approximately 62% accuracy in channel coding recognition and 64% accuracy in interleaving recognition, while the convolutional neural network (CNN) achieves 38% and 65% accuracy, respectively. However, when the ratio shifts to 40:60, all recognition models exhibit high accuracy, exceeding 70%. Therefore, the proposed models demonstrate that the recognition performance improves as the difference in the ratio of Os to 1s in the message increases.