Research On Key Techniques For Communication Emitter Identification

Communication emitter identification refers to the technology of individual emitter identification using only the external characteristics of communication signals.It mainly includes two aspects: signal fingerprint feature extraction and classification identification.The signal fingerprint feature extraction is the core of communication emitter identification.However,most of the current research on the mechanism of fingerprint feature generation of communication emitter signal is still in the theoretical level.The extracted fingerprint features of communication emitters are not effective and stable,and need many priori conditions,so it is difficult to adapt to the actual changeable signals.Moreover,the traditional fingerprint feature extraction methods are mainly manual,which leads to the time-consuming and laborious feature extraction and the results are difficult to guarantee,thus making the communication emitter identification still face more problems in practical engineering applications.Signal time-frequency spectrum can effectively reveal the characteristics of the signal in the time-frequency domain.The singular value and singular vector of signal time-frequency spectrum can effectively extract the fingerprint characteristics of the signal,and has a strong adaptability to the signal,so it has a good practicability.One of the main contents of in-depth learning research is how to make the machine learn the input data,mine out the regular information,and realize the automatic extraction of the features which can effectively describe the input data.It requires less background information,the nature of mathematical physics and knowledge of the problem itself,and it is difficult to make the effect of time-consuming and laborious.Guaranteed feature extraction is liberated from work.In this paper,depth learning theory is introduced into the field of specific emitter identification,and the powerful feature representation and data mining capability of depth learning are used to learn the input emitter signal data independently,so as to realize the adaptive intelligent extraction of effective fingerprint features of the input emitter signal data.Finally,the fingerprint features extracted by depth neural network are classified and identified,and the feature database is built and updated to form an intelligent identification system for communication emitters.The research contents are as follows:(1)The research status of fingerprint characteristics of communication signals is systematically summarized.The research work of fingerprint characteristics at home and abroad is analyzed from three aspects: fingerprint mechanism,main extraction methods and classification and identification methods.The future application prospects of communication emitter identification technology in military and civil fields are discussed.(2)The basic concepts and main algorithm flow of time-frequency spectrum singular value and singular vector are introduced.The characteristics of time-frequency spectrum singular value and singular vector are analyzed,and the Gaussian mixture model classifier is introduced.Finally,an experimental data acquisition system is designed.The validity and stability of the singular value and singular vector are analyzed by the experimental data.The system has good performance.(3)two identification methods of signal emitter signals based on deep learning are proposed.One is that the time-frequency graph of the signal is used as input to realize the identification of communication emitter through CNN network,the other is that the time series of the signal is input to LSTM network to realize the identification of communication emitter.(4)Design and implement the fingerprint feature analysis test system of communication signal,including the design of signal receiving and acquisition system and data processing software,which integrates the functions of receiving communication signal,signal data preprocessing,fingerprint feature extraction and individual classification and identification.Then the fingerprint feature extraction and analysis of the field measured target data are carried out by the test system.The results verify the effectiveness of the proposed method and the practicability of the test system.