Research On Spectrum Anomaly Detection Technology For Management Of Radio

As a national strategic resource,radio spectrum resources play a vital role in the civilian and military fields.Therefore,all countries attach great importance to spectrum resource management.The detection of the abnormal state of the spectrum is one of the most important contents of spectrum resource management and monitoring.The traditional method is to establish a detection model through a supervised algorithm,and the types of abnormal signals that can be detected are single.In actual scenarios,the probability of signal abnormality is relatively small and the reasons are numerous,and the existing model methods cannot be adapted.Therefore,in response to the abovementioned problems,based on unsupervised ideas and artificial intelligence methods,this thesis has launched a research on the detection methods of three kinds of spectrum abnormalities: malicious electromagnetic interference,occurrence of unauthorized signals,and illegal use of authorized signals.Specifically,the main work content of the thesis includes the following four aspects:(1)Build a spectrum anomaly model,and build a data acquisition platform to collect the data set needed for the experiment.Based on the comparative analysis of related spectrum anomaly models,model the abnormal state of the spectrum according to actual needs.Then according to the spectrum anomaly model,the electromagnetic signal data set and frequency scanning data acquisition platform are built,and the data set required for subsequent algorithm verification is constructed.(2)A detection algorithm based on relative wavelet time entropy is proposed to solve the problem of detecting abnormal state of spectrum caused by malicious electromagnetic interference.The algorithm calculates the relative wavelet time entropy curve of two normal working signals,observes its fluctuation range to determine the detection threshold.Then the relative wavelet time entropy curve of the signal under test and the normal signal is calculated,and the spectrum anomaly detection is achieved by comparing with the detection threshold.Finally,the FM electromagnetic signal data set is used for simulation experiments,and the results show that the algorithm is effective and feasible.(3)A detection scheme based on a generative countermeasure network is constructed,which solves the problem of detecting the abnormal state of the spectrum caused by the occurrence of unauthorized signals.The scheme is to input the timefrequency diagram of the normal working signal into the neural network,and establish the discriminant model by learning the statistical distribution law of the frequency spectrum,and use the saved model to realize the spectrum anomaly detection.Finally,the electromagnetic signal data set is used to carry out a simulation test,which improves the detection accuracy to 97.1%,and according to the comparative test,it is verified that the detection performance of this scheme is better than the detection scheme based on relative wavelet time entropy.(4)An anomaly detection scheme based on long-short term memory network(LSTM)spectrum prediction is constructed,which solves the problem of detecting the abnormal state of the spectrum caused by the illegal use of authorized electromagnetic signals.The scheme is based on the frequency usage rules of authorized electromagnetic signals,and by using the periodicity and predictability of the channel occupancy sequence,With the help of LSTM,the channel occupancy sequence of the next time period is predicted,and the difference between the predicted result and the actual result is compared to judge the abnormality.Finally,a simulation analysis of the frequency band scan data set shows the effectiveness of the scheme.The thesis mainly studies the detection of various spectrum anomalies in the electromagnetic spectrum management and control process in the absence of anomalous signal data sets.Through the solution proposed in the thesis,the detection of the abnormal state of the spectrum under the unsupervised state is effectively realized.