Study On Open Set Communication Radiation Source Identification

Communication radiation source identification is a key technology in Information Security and Electronic Countermeasures.The accuracy of communication radiation source identification affects the level of the whole information countermeasures system.Therefore,it is of great significance to study the method of communication radiation source identification and improve the accuracy of radiation source identification.First,a Weibull-calibrated Support Vector Machine(W-SVM)based open set identification algorithm for multiple types of communication radiant sources was proposed in this thesis,which generates a Compact Abating Probability(CAP)model from the basic 1-class SVM(OCSVM),converts the modeling of the unknown class into a probability value that decaying with distance,and identifies the unknown class by setting a probability threshold.Based on Binary SVM and Statistical Extremum Theory,weibull distribution and reverse weibull distribution were used to calculate the probability estimation of the sample of from positive sample and not from negative sample respectively,and the classification of the sample was further determined.Secondly,a closed set communication radiation source recognition method based on deep learning is proposed.The Short-Time Fourier Transform(STFT)is used to extract the subtle features of the original data.And the shallow and deep features of the communication radiation source signals are extracted by Two-Channel Convolution Neural Networks(TCNN).And then a combined loss function combining MI divergence and Softmax is proposed to realize closed set recognition of communication radiation source signalFinally,an open-set communication radiation source identification method based on deep learning is proposed.On the basis of the closed set recognition algorithm,the fingerprint feature of communication radiation source is extracted by TCNN.In the training phase,the distance distribution of each type of data is calculated and the threshold value is set.In the test phase,the distance between the input sample and the center of each known class is calculated.When the distance between the input sample and the center of each class exceeds the threshold value,the input sample is judged as an unknown communication radiation source class.The methods proposed in this thesis improve the recognition accuracy of closed set,realizes the open set recognition of unknown communication radiation source signal,and provides a new idea for the study of communication radiation source recognition.