Spectrun Prediction Method For ISM Bands

Spectrum prediction technology is an important part of cognitive communication.The spectrum prediction technology can effectively use the internal correlation between the spectrum occupation statistics,infer the occupation status of the radio spectrum in the future from the known spectrum occupation,so as to assist the spectrum sensing process and improve the spectrum utilization.ISM(Industrial Scientific Medical)frequency band for industrial,scientific and medical use is a spectrum resource that can be used without authorization.Wireless LAN,Bluetooth,ZigBee and other wireless networks all work in this frequency band.Due to the time-varying,dynamic and high burst characteristics of the spectrum state of ISM frequency band,it is difficult to discover the spectrum occupation pattern of ISM frequency band,which is the difficulty of spectrum prediction.Because the Wi-Fi network based on IEEE 802.11 standard is the most common network technology in ISM frequency band,this thesis mainly studies the spectrum prediction of ISM frequency band based on Wi Fi network.The main work and innovation of this thesis are as follows:1.This thesis first introduces the existing spectrum modeling and spectrum prediction technology,and introduces the ISM frequency band and analyzes the spectrum characteristics.Then,the necessary significance and difficulties of spectrum prediction in ISM band are analyzed.Through the research and analysis of the existing spectrum prediction methods in ISM frequency band,it lays the foundation for the subsequent improvement of the algorithm of spectrum prediction in ISM frequency band.2.In this thesis,we use the Markov modulated Poisson process(MMPP)to model the frequency spectrum of Wi-Fi network,and use MMPP/G/1/K model to simulate the AP(Access Point)activity in Wi-Fi network.In this model,the dynamic and burst characteristics of Wi-Fi traffic are considered,and the frequency,duration and probability distribution of the blank in Wi-Fi frequency band are described mathematically.According to the results of model analysis,the frequency and duration of Wi-Fi blank are enough to allow meaningful cognitive communication,which lays a foundation for the subsequent spectrum prediction.3.In order to improve the spectrum prediction accuracy of ISM frequency band;this thesis proposes a spectrum prediction algorithm of K-LSTM(K-means Long and Short Term Memory)network.The algorithm classifies the Wi-Fi spectrum according to the characteristics of the spectrum data,and forecasts the spectrum based on the classification results.It overcomes the problem of poor generalization of the traditional LSTM network model in the multi-mode Wi-Fi spectrum,so as to improve the spectrum prediction accuracy to a certain extent.The simulation results show that the proposed K-LSTM algorithm can significantly improve the prediction accuracy of complex spectrum mode compared with the existing algorithm.In this thesis,RBF(Radial Basis Function)network with local response is also used for spectrum prediction.In the multi-mode Wi-Fi spectrum data,the local response characteristics of the network can improve the generalization of the prediction model.However,due to the high computational complexity of the network,it is not suitable for the prediction of a large number of data,and can not make full use of the advantages of the network,so the performance of the network is not as good as the K-LSTM algorithm.