Modulation recognition of communication signals based on deep learning

Deep residual shrinkage network was used to realize modulation signals feature extraction and classification recognition according to the characteristics of signals to further improve the accuracy of communication signals modulation recognition under low signal-to-noise ratio. The deep residual shrinkage network can extract feature from noised signals more effectively with the help of soft thresholding under the attention mechanism, and its modulation recognition rate especially under low signal-to-noise ratio compared with convolutional neural network and deep residual network has improved significantly through experiments. The network pruning method was used to optimize the model to improve the training efficiency of the deep residual shrinkage network model. The method effectively compresses and accelerates the model while ensuring that the model still has good recognition accuracy.