A New Modulation Recognition Method Based on Flying Fish Swarm Algorithm
Jingpeng Gao,Xu Wang,Ruowu Wu,Xiong Xu
DOI: https://doi.org/10.1109/access.2021.3079131
IF: 3.9
2021-01-01
IEEE Access
Abstract:The modulation recognition method based on deep learning plays a significant role in the intelligent communication system. To further improve the recognition rate, especially in the case of small samples with a low signal-to-noise ratio, this paper proposes a new modulation recognition method based on flying fish swarm algorithm. First, Short-Time Fourier Transform, Choi-Williams Distribution, and Cyclic Spectrum are combined to complete multi-channel signal processing. Second, AlexNet, VGGNet, GoogLeNet, and ResNet are transferred to realize feature extraction. Third, the support vector machine classifies the modulations after dimension reduction and feature fusion. Finally, the flying fish swarm is proposed to optimize the signal processing methods, the types of networks, the layers of networks, the dimensions of features, and the parameters of the support vector machine. The method can accurately recognize BPSK, QPSK, OQPSK, 8PSK, 4ASK, QAM16, QAM32, and QAM64. The simulation results show that the average recognition rate of modulation is 94.5% at SNR of 0 dB and 84.7% at SNR of −4 dB. Besides, the proposed modulation recognition method possesses good robustness under low SNR conditions.
computer science, information systems,telecommunications,engineering, electrical & electronic