Transfer Learninng in Polytime Codes Signal Recognition
Lu Gao, Peng Qin, Hu Li, Shangyue Wang, Heng Sun, Yi Lu · 2019
Aiming at the problem that existing methods of radar signal modulation recognition do not perform well under low SNR, this paper proposes a new polytime codes signal modulation recognition algorithm based on the transfer learning. The paper transforms the polytime codes signal into time-frequency image so that CNN can be used to extract features. According to the theory of transfer learning, the paper chooses VGG-19 to extract feature, while the classifier is the Probabilistic Neural Networks (PNN). What's more, to satisfy the input requirement of the VGG-19, the Stacked AutoEncoder (SAE) can help to reshape the size of T-F image without losing essential information. The simulation results show that the recognition rate of the proposed algorithm reaches at 97% at 3 dB. Thus the proposed algorithm has a good application in the field of electronic battlefield.