DenseNet-ResNet-LSTM model for modulation recognition of communication signal

Abstract

The traditional artificial neural network provides a low recognition rate in the modulation recognition of communication signals, suffering from the difficulty of feature extraction. Also, it requires a high signal-to-noise ratio (SNR). In order to solve these problems, this paper proposes the combined model: DenseNet-ResNet-LSTM. In the proposed model, DenseNet and ResNet extract different spatial features of samples, and then LSTM extracts the sequence of samples. Also, the attention mechanism is employed to improve the learning efficiency and ability to learn important features. Experimental results show that the proposed model achieves higher accuracy and better generalization ability over the CNN-LSTM network.