| Target recognition of emitter signal is to measure the characteristics of the received modulated signal and to determine the individual that generates the signal according to the information obtained.Emitter signal target recognition has become important development direction in modern electronic countermeasure field,domestic study of this technology,the equipment cost is high,the feature extraction is not ideal,such problems as classification recognition model is not accurate,resulting in 0 ? 20 d B SNR environment,the classification of the emitter signal recognition rate is not high.Therefore,this paper puts forward the design of target recognition system for radiator signal by using walkie-talkie as research object.In this paper,the mechanism of the generation of subtle features of radiation sources and the traditional feature extraction methods are understood.Then the research of the method in this paper is carried out: according to the collected signals of walkie-talkie,the pretreatment work such as filtering and sample division is carried out.Because of the traditional feature extraction of the proposed feature contains incomplete information and characteristics of signal dimension is overmuch,may affect the accuracy of recognition,in order to avoid this problem,based on the microscopic characteristics of mechanism to extract the five characteristics of the entropy to describe the signal may produce complex changes in values,using the mean and standard deviation analysis of characteristic of validity,Then extract the five characteristics of the fractal dimension to describe the signal on the space shape of complex changes,from the perspective of two different well reflects the microscopic characteristics of changes in the signal,are a good characterization of the differences between different signals,Relief F algorithm finally eliminated the original little useful for identifying or redundant features of each other,The remaining features constitute the final eigenvector.The recognition rate of emitter signal target recognition depends not only on the extraction of feature but also on the selection of classification model.This paper mainly uses machine learning method to build classification model,chooses decision tree,support vector machine and extreme learning machine to build three algorithms,and makes corresponding optimization of the algorithm according to the signal extraction feature quantity.In the construction of the decision tree classification model,ID3 was selected as the main algorithm to construct the decision tree.The minimum sample number of the leaf nodes of the decision tree was selected through cyclic comparison to construct the decision tree model and carry out classification,recognition and analysis.In the construction of the support vector machine classification model,the genetic algorithm was selected to optimize the penalty coefficient and kernel function parameters,and the optimized classification and recognition model was constructed,and the classification and recognition analysis of the classification model was carried out.In the construction of the classification model of Extreme Learning Machine,the cloud adaptive particle swarm optimization algorithm was proposed to optimize ELM,and the superiority of the algorithm was verified by parameter optimization.The optimized classification and recognition model was constructed,and the classification and recognition analysis of the classification model was conducted.Through experimental comparison,the performance index and recognition rate of the classification model optimized by cloud particle swarm optimization are the best.Finally,the process of collecting,preprocessing feature extraction and classification model construction is realized with the interphone signal as the object.The software algorithm is completed by MATLAB software.At the same time,in order to make the user have a more intuitive experience,a graphical user interface is designed to show the whole process of recognition. |
PDF Full Text Request