Modulation Recognition of Communication Signals Based on Deep Learning Joint Model

Abstract

In modern communication, it is often required for a non-cooperative party to identify the modulation mode when no prior knowledge is given to facilitate the subsequent demodulation and analysis. However, the traditional modulation recognition process requires cumbersome and uncertain manual signal-feature extraction, making it inapplicable to the complex communication environment. In order to overcome this limitation, this paper proposes a communication-signal modulation recognition model based on the dense connection network (DenseNet) and residual connection network (ResNet). The convolutional block attention mechanism (CBAM) is introduced into the DenseNet and ResNet structures, significantly enhancing the modulation recognition accuracy of the proposed global network model. Besides DenseNet and ResNet, the long short-time memory (LSTM) network is also adopted. The experimental results show the performance of the proposed model.