Automatic Modulation Classification and SNR Estimation Based on CNN in Physical-layer Network Coding

In this paper, we first propose the Automatic Modulation Classification (AMC) problem based on the Physical layer Network Coding (PNC) system and elaborate in detail. We use Convolutional Neural Networks (CNN) to identify nine cases including three modulation formats with three phase shifts respectively, and estimate the Signal-to-Noise Ratio (SNR) simultaneously. As the result, we correctly identify several modulation formats and typical phase offsets with a 100% recognition rate, and estimate the received signal-to-noise ratio effectively with recognition rate above 98%.