Modulation Recognition based on Complex Binarized Neural Network
Jingyan Wang, Qi Lin, Yu Lai Han · 2022 9th International Conference on Dependable Systems and Their Applications (DSA) · 2022
In recent years, communication technology is developing rapidly. With an increasing number of radio devices entering people's life there have been emerged much more communication modulation methods being used when modulation recognition technology has very important application value. In the past few years, the development and progress of deep learning is remarkable, using deep learning algorithms to solve the modulation recognition problem also has an excellent performance. Faced with massive amounts of electromagnetic data and resource limited hardware, existing technologies cannot meet the needs of deployable systems to handle short-term observations or the sudden arrival of a short-term signal and immediate decision-making. In this paper, Binary Complex Neural Network (BCNN) combining Binary Neural Network (BNN) and Deep Complex Neural Network (DCN) is used for modulation recognition, and we propose a scheme of Lightweight Neural Network for Electromagnetic Signal (LNN-ES). Experiments show that compared with BNN, BCNN has a better result in the recognition accuracy. And compared with the residual network, the lightweight neural network proposed in this paper can achieve the same recognition accuracy with less parameters.