Modulation Recognition Based on Neural Network Ensembles

This paper studies the modulation recognition of digital communication signals based on neural networks. The BP neural network ensembles method is put forward, which is a linear composition of the BP neural networks. The recognition accuracy of ten different modulation formats is given according to the model above in feature extraction. The approach presented is superior to a neural network algorithm in existing articles. The result shows that the method proposed can recognize complex signal modulation formats availably. The overall recognition accuracy is basically up to 100% in the sample data of this paper when the SNR is more than 8 dB.