In recent years, more and more researchers have tried to introduce the recognition algorithm based on deep learning in the field of communication emitter individual recognition, and have made some achievements in the ideal experimental environment where the signal parameters are known and the samples to be recognized are pure.
However, in the real scene, due to the change of signal receiving environment, the change of receiving system mode and other factors, the data to be identified is uncertain, and the signal-to-noise ratio of signal samples becomes low, which is a typical evolution scene with uncertain data.
Most deep learning models are difficult to extract discriminative features from the above data evolution scenarios, and the recognition results will become unreliable.
Therefore, the emitter individual recognition algorithm based on deep learning is still difficult to replace the traditional signal analysis algorithm relying on expert experience, and the former appears as an auxiliary classifier in practical application .
The overall architecture of deep residual shrinkage networks is shown in Figure 2.1. DRSN is mainly divided into three modules: signal denoising module, residual shrinkage filtering module and full connection dimension reduction module.
Deep residual shrinkage network [2-3] is a deep learning method originally used in the field of fault diagnosis, which is mainly suitable for the case of strong noise.
 Song Yuxuan. Research on evolutionary deep learning and application of communication signal recognition [D]. Xi’an University of Electronic Science and technology, 2020
 M. Zhao, S. Zhong, X. Fu, B. Tang, M. Pecht, Deep residual shrinkage networks for fault diagnosis, IEEE Transactions on Industrial Informatics, vol. 16, no. 7, pp. 4681-4690, 2020.
 Code: https://github.com/zhao62/Dee…