Research On Specific Communication Emitter Identification Based On Deep Learning

Specific emitter identification refers to the process of extracting characteristics of received signals and identify the emitter individual based on the characteristics.It is of great significance for improving the security of wireless communication systems and improving the capabilities of military electronic countermeasures.The traditional specific emitter identification method first designs a special manual feature extraction method to extract the features,and then uses a classifier to identify it.This method has disadvantages such as high feature extraction complexity and limited application range.With the development of artificial intelligence,deep learning has made many new advances and breakthroughs in the research of specific emitter identification.The use of deep neural networks can obtain the ability to extract features through learning,so there is no need to design a special manual feature extraction method.Compared with traditional methods,deep learning methods not only have a wider range of applications,but the algorithm’s performance is often better.Therefore,this thesis uses deep learning technology to carry out research on specific emitter identification.The main research work and innovations include:In order to fully extract features from the signal,a specific emitter identification method based on signal graph and ensemble neural network is proposed.By constructing the signal sequence as a signal graph,the signal sequence can be transformed from Euclidean space to non-Euclidean space.After that,an ensemble neural network was further designed to extract sequence features and graph features to obtain rich feature information.The experimental results on the self-collected data set and the open source dataset both show that the specific emitter identification performance of this method is better than traditional machine learning based methods and the existing deep learning based methods.This thesis explores the influence of the attention mechanism on the neural network’s specific emitter identification performance,as well as the problem of specific emitter identification under unbalanced samples scenario.Different sample imbalance solutions are introduced at the data level and the loss function level respectively.The experimental results show that after using the sample imbalance solution,the algorithm’s emitter identification performance in the sample imbalance scenario can be significantly improved.