Automatic Modulation Recognition Based on Multi-Channel Neural Network Model
Xianchao Zhang, Shengyu Ma, Jian Shi, Panpan Li, Guangxue Yue · 2022
Aiming at the problem of low accuracy of existing wireless communication blind recognition methods, a novel multi-channel deep learning framework based on the Convolutional Long Short-Term Memory Fully Connected Deep Neural Network (MC-CLDNN) is proposed. We fully combine the advantages of convolution neural network (CNN), gated recurrent unit (GRU) and deep neural network (DNN) in feature extraction ability to improve the efficiency of network training. Furthermore, to alleviate the problem of gradient disappearance in the network training and reduce the negative effect of pooling layer processes time series data on the subsequent sequence model, the skip connection is added to the network model. We verify the feasibility of the model based on opensource dataset RadioML2016.10a. The simulation results show that the proposed model can identify most modulation modes effectively, and has the characteristics of high recognition accuracy and strong generalization ability.