Multiple-Stream Model with CNN-LSTM for Automatic Modulation Recognition

This paper presents a multiple-stream deep learning (DL) model for automatic modulation recognition (AMR). The proposed model crosses convolutional neural network (CNN) and long short term memory (LSTM) to extract features from different combinations of representations of received signals and make improvements from GaussianDropout. The results show that the proposed model compared with existing models has higher recognition accuracy for various modulation.