Modulation recognition using an FPGA-based convolutional neural network

As wireless communications continues to gain in importance and spectrum traffic becomes more and more dense, the challenge of fast modulation classification is also increasing. Although modulation classification has been a well-studied problem, achieving high accuracy from a small number of samples is difficult. Moreover, most methods are not designed for line-speed operation over multiple channels or bands. In this paper, a deep learning model using a constellation diagram approach and its FPGA implementation are applied to solve this problem.