Automatic Modulation Recognition of Communication Signal Based on Deconvolution Generation Adversarial Network
Zhenyu Wang, Fan Zhou, Xiao Han, Lan Zhang · 2024
To solve the problem of low signal recognition accuracy and insufficient network training under the few-shot condition, this paper proposes a generative adversarial network modulation recognition algorithm based on deconvolution feature reconstruction, which uses a small number of labeled samples to generate signal fake samples satisfying convolutional neural networks. The simulation results show that under the few-shot condition, the average recognition accuracy of the proposed algorithm can be improved by 9.3% compared with the traditional convolutional neural network algorithm when the SNR ranges from -6 to 6dB. Compared with the traditional generative adversarial network algorithm, the average recognition accuracy can be improved by 2.13%, effectively realizing the signal modulation recognition under the few-shot condition.