Cyclic Cumulant Based on Communication Signal Multilayer Neural Network Modulation Pattern Recognition

Abstract

After we deduces and simulates the cyclic spectrum theory of MPSK and MQAM modulated signals. The cyclic spectrum of the modulated signal is preprocessed and the original feature set is formed as the input data of the CNN. Feature extraction of cyclic spectrum from convolution and subsampling layers of CNN, so that the modulation signal can be effectively identified within and between classes. The experimental simulation results show that the classification and recognition algorithm of communication signal modulation mode proposed in this paper has better modulation recognition accuracy than other traditional modulation recognition algorithms in the case of low SNR. And when the signal to noise ratio is lower than 5db and higher than -5db, the modulation pattern recognition accuracy can be reached up to 92%.