Optimized Deep Learning Technique Based Digital Modulation Identification

Abstract

In many forms of wireless communication, such as cognitive radio and signal recon, modulation detection is a crucial task. The demand for accurate identification of OFDM signal has increased due to the diversity of modulation techniques and the complex route environment. The enhanced large data processing and categorization capabilities of D-L are thought to be viable answer to these issues. In this study, an effective deep neural network (DNN)-based technique for digital modulation identification is proposed. In addition, we introduce the particle-swarm-optimization (P-S-O) approach to improve the amount of concealed layer nodes of the DNN in order to address the issue that the original D-N-N is stuck at local minimum values and the amount of concealed layer nodes needs to be explicitly selected.