Communication transmitter individual feature extraction method based on stacked denoising autoencoders under small sample prerequisite

For the goal of improving the effectiveness and robustness of identifying communicatioin radio individual under small sample prerequisite, we proposed a communication radio individual identification method based on stacked denoising auto-encoder, which can learn fingerprint feature from input unlabeled communication radio signal samples and then extract low-dimension feature of raio signals facing to high-dimension sample space. Finally, the accurate individual identification phase will be accomplished in classifier. Experiments showed the effectiveness and reliability of method proposed on three different communication radio data set.