Personal Identification Of Communication Radiation Source Based On Deep Belief Network

The battlefield environment is complex and variable,and determining the level of the target radiation source plays an important role inThe battlefield environment is complex and variable,and determining the level of the target accurately targeting the target.The subtle features of the communication radiation source are able to distinguish the basis of the individual communicating the radiation source.However,the traditional method of individual identification of radiation source is based on the preset determination formula or under the corresponding preset conditions,and the a priori requirement for the signal is relatively high,and the applicable range is relatively fixed.And distinguishing the feature vector of the communication radiation source individual is complicated.In view of the above problems,this thesis studies the three aspects of the selection of the individual characteristics of the communication radiation source,the establishment of the individual identification model of the radiation source based on the deep belief network and the optimization of the individual identification model of the radiation source based on the deep belief network.The main contributions of the paper are as follows:By comparing the individual characteristics of the communication radiation source,the characteristics of the intermodulation interference signal in the communication radiation source signal band,the carrier frequency characteristics of the communication radiation source and the modulation parameter characteristics of the communication radiation source are selected as the basis for the individual identification of the communication radiation source.A radiation source individual identification model based on deep belief network is established.The radiation source signal is preprocessed to obtain a rectangular integral bispectrum of the communication radiation source signal,and each restricted Boltzmann machine is trained from the bottom up using a high-order spectrum,and multiple weights are obtained to obtain appropriate weights,hidden layer deviations and visible The deviation of the layers,thereby extracting the intermodulation interference signal characteristics of the radiationclassifier,and a deep learning network for individual identification of communication radiation sources is obtained.3.Optimize the above model.The model is optimized from several parameter settings such as pre-training learning rate,network layer number and network pre-training times.4.Using the optimized network to classify and identify the carrier frequency characteristics,modulation parameters and intermodulation interference characteristics of the communication radiation source,and compare the identified results.The results show that the method can effectively identify the individual sources of communication radiation under less a priori conditions.And through simulation analysis,it is found that the pre-trained learning rate is set to 0.01,the network layer is set to 4 layers,and the pre-training times are set to 20 times to obtain an ideal network model.At the same time,through simulation analysis,it is found that the recognition rate of intermodulation interference characteristics is better than that of communication radiation source and the modulation parameters of communication radiation source.It is suitable as the basis for subtle feature recognition of communication radiation sources.