The Research On Correlation Recognition Of Signal And Emitter Based On Machine Learning

In the increasingly complex electromagnetic environment,the correlation recognition of signal and emitter has attracted wide attention of scholars at home and abroad.The identification of signals and emitter refers to the establishment of the relationship between signals and emitter,and the identification of emitter signals can effectively identify the corresponding emitter sources.Firstly,the identification problem of signals and emitter sources is divided into two parts: association recognition of signals and emitter sources,and identification of signals and individual emitter.The former realizes the recognition of emitter type by identifying different types of communication,radar and other emitter signals,and according to the oneto-one correspondence between signals and emitters.The latter recognizes the different emitter source individuals after identifying the emitter type,that is,identifying the signals of different emitter individuals of the same type and working mode,and then recognizing the individual of the emitter source according to the one-to-one correspondence between the signal and the individual of the emitter source.Secondly,aiming at the problem of correlation recognition between signal and emitter type,this paper takes the signal samples with known emitter type as the research foundation,and associates electromagnetic signals with different types of communication,radar and other emitters,and then realizes the recognition of emitter type by establishing signal classifier.In practical application,according to the output category of unknown signal in classifier,the recognition of emitter type can be realized.Because of the inconsistency of the different types of emitter signals,the traditional emitter recognition methods can't extract uniform features for different types of emitters.In this paper,the deep learning idea is introduced,and the convolution neural network is used to extract the signal features intelligently.The problem of identifying different types of emitter signals is unified,which avoids the shortcomings of indistinct feature characterization and insufficient number of features caused by artificial selection of features.Thirdly,aiming at the problem of individual correlation recognition between signals and emitters,the individual of different emitters of the emitter type is classified and identified after the effective recognition of the emitter type.Based on the signal samples with known individual information of the emitter,the signal is correlated with different emitter individuals of the same type and working mode,and the individual recognition of the emitter is realized by establishing a signal classifier.In practical application,according to the output category of unknown signals in the classifier,the individual emitter can be recognized.In this paper,firstly,the classification and recognition method based on bispectrum and machine learning algorithm is used.By calculating the integral bispectrum of emitter signal and calculating the peak value as the signal feature,the support vector machine is used to classify and recognize the signal feature.The simulation results show that this method has good classification performance for four intercoms of the same type,but it has some limitations when the number of intercoms to be identified is large.Finally,in view of the shortcomings of bispectrum and support vector machine,the deep learning convolution neural network is introduced into the individual correlation recognition of signal and emitter.The convolution network is used to extract the signal features directly,and the back propagation is used to modify the network parameters to realize the intelligent extraction of the signal features.In this paper,the performance of the algorithm is tested on ten intercoms of the same type.The simulation results show that the algorithm has good recognition accuracy and practicability.