Research On Individual Identification Method Of Satellite Communication Emitter Based On Expert Knowledge

Individual identification of communication radiation sources refers to the process of target recognition for radiation source individuals based on the individual unique characteristic information(also known as radio frequency fingerprint)formed by communication radiation sources due to manufacturing process,accumulated damage during use and other reasons.With the continuous progress of space technology and the vigorous development of satellite industry in recent years,space reconnaissance and satellite communication countermeasures have also made significant progress among the world’s military powers,and satellite communication radiation source identification has become a current research hotspot.The identification method based on expert knowledge can break through the limitation of incomplete samples,and has the advantages of high efficiency and strong interpretability,which is an important branch in the field of communication radiation source identification.Individual identification technology for satellite communication radiation sources has important applications in electronic countermeasures,security authentication assistance,and other fields.Currently,there is a lack of dedicated identification methods for individual identification of satellite communication radiation sources.Due to the complex fingerprint feature composition within satellite communication equipment,most identification methods based on traditional single features and classifiers can only be used in a certain type of communication scenarios,with poor applicability.To address these issues,this paper introduces a feature fusion method and a classifier integration architecture into the individual identification of satellite communication emitters,building a feature library and decision level for satellite communication scenarios,and ultimately achieving multi-scene adaptive emitter individual identification for satellite communication.The main work of this paper is as follows:1.Based on the radiation source fingerprint formation principle of satellite communication system,the effective mapping features are extracted using expert knowledge and preliminary verification methods to build the features library.Starting from the formation mechanism of oscillator deviation,quadrature modulation distortion,power amplifier nonlinearity,we extract wavelet features,higher-order spectrum features,fractal features,which can map the differences of the working modules,and test the equipment data sets with similar communication mechanism to verify the effectiveness of the features.2.Aiming at the problem that a single feature can’t comprehensively map the individual difference characteristics of satellite communication radiation sources,a feature fusion method is introduced to improve the mapping effect and recognition accuracy.In this paper,canonical correlation analysis and discriminant correlation analysis are used to fuse multi-dimensional features,thus greatly improving the mapping ability of complex modules of satellite communication equipment,so that the fingerprint features of multiple modules can be used for training and improve the recognition accuracy.Also the nonlinear dimension reduction method is introduced to replace the traditional linear dimension reduction method to better improve the discrimination of nonlinear features and enhance its linear separability.By introducing new feature processing methods,an individual emitter identification method for satellite communication scenarios is proposed.3.Aiming at the problem of poor robustness of traditional single classifier,the classifier integration architecture is introduced to improve the identification effect.In this paper,we propose to use Adaboost static integration and Stacking dynamic integration as decision-level identification methods,which combine multiple weak classifiers of poor robustness into strong classifiers after training to compensate for the identification error on some outlier sample points.Also we propose a dynamic classifier selection algorithm based on clustering,which makes dynamic selection of classifier architecture on each sample point,and obtains the optimal architecture weight,improving the recognition accuracy.4.GUI interface is designed through Matlab to realize preliminary application.The recognition interface is designed by using Matlab toolbox,which is convenient to guide the selection of mapping features and discrimination architecture based on expert knowledge,and display the recognition results.