Research On Short-wave Frequency Prediction Optimization Algorithm Based On HMM

With the development of the adaptive electronic warfare technologies,electronic attack is drawing much more attention.It is essential to predict the behavior of enemy signals if both sides want to gain information superiority in electronic attack.Working frequency of signal is one of the main behaviors,so it is of great significant to predict the working frequency.In the wartime environment,short-wave signal has unique advantages and occupies an important position in military communication.Therefore,this thesis studies the prediction of working frequency of short-wave signal.Since the limitation of the existing short-wave signal frequency prediction methods and the excessive human factors.This thesis studies a new short-wave signal frequency prediction method based on hidden markov,and then conducts simulation verification of this method.Firstly,this thesis introduces and analyzes the existing short-wave signal frequency prediction methods,and points out their disadvantages.The existing short-wave frequency prediction methods are all indirect prediction of short-wave signal frequency,and the selected data sets were ionospheric data observed in early years or simulated data of frequency prediction software,which have certain limitations.Based on these problems,this paper studies the optimization of operating frequency prediction algorithm for short-wave signals based on hidden Markov.Secondly,in view of the prediction problem of data set,we consider the change of the ionosphere has a direct impact on the short-wave communication channel characteristic parameters change.Therefore,this thesis analyzes the characteristics of short-wave communication channel,points out the change of channel characteristic parameter range and then we can choice of gauss scattering gain tap delay line(Watterson)channel model for corresponding improvement by analyzing shortwave channel model.This improvement can achieve shortwave signal working frequency jump and simulate prediction data sets.According to the different value ranges of impact factors under normal and special circumstances,the data of two sets of 10000 points are simulated to provide data sets for the following prediction.Finally,the feasibility analysis was carried out on the hidden markov,and establish the working frequency of short-wave signals based on hidden markov prediction model.Using the data obtained from the above set to calibrate the model,and considering the phenomenon that there may be individual combinations of influencing factors in the data set that do not exist,the "zero probability" problem.In this thesis,the "zero probability" problem is solved by adding one smoothing,and the simulation analysis and performance comparison are carried out.Mean Relative Error,Mean Absolute Deviation and Max Absolute Relative Error are used as the basis of performance analysis.The results show that the prediction accuracy of the basic algorithm can reach 81.55% under normal conditions,and the prediction accuracy of the optimized algorithm can reach 89.62%.Under special circumstances,the prediction accuracy of the basic algorithm can reach 75.83%,while the prediction accuracy of the optimized algorithm has been improved by 8.56%.This method is feasible to some extent.