FPGA Deployment of Deep Neural Network for Modulation Recognition

Modulation recognition is an important task in intelligent communications, and modulation recognition using deep neural network has gained considerable attention due to the rapid advancement of deep learning technology. However, implementing modulation recognition using deep neural network on a general-purpose platform is time-consuming, which makes it difficult to meet the high-speed and real-time requirements of practical applications. To address the issue, this paper proposes a field-programmable gate array (FPGA) deployment system for modulation recognition network. The proposed system leverages parallel computing and pipeline methods to significantly enhance the inference speed of the modulation network, Experiment results show that FPGA as a hardware accelerator can run parallel calculations up to 139 times faster than the ARM Cortex-A9 processor.