Research On The Spectrum Prediction Algorithm In Cognitive Radio System

Cognitive radio (CR) is an intelligent wireless communication system that can be aware of its environment and adapt to the input stimuli, and has been intensively studied as a promising strategy to solve the scarcity of spectrum. It can obtain the access of licensed spectrum which is unoccupied or partially occupied as a secondary user (SU). In cognitive radio system, spectrum prediction methods are able to gain the usage pattern and predict the spectrum holes by analyzing the spectrum. The information can be used by cognitive devices to realize intelligent spectrum sensing and dynamic spectrum access, which result in the improvement of throughput and QoS of cognitive radio system.In this thesis, we mainly focus on the spectrum prediction methods and its performance in cognitive radio. Firstly we have a survey of current spectrum prediction methods. According to the modeling problem of license spectrum, we proposed a spectrum model based on M/M/N queuing model. The parameters estimating method and correspond theory analysis are given sequentially. In order to balance between the complexity and accuracy of current spectrum prediction methods, back propagation neural network is proposed to learn and predict the spectrum holes in this thesis. We use the measured data collected by Agilent spectrum analyzer. Simulation and numerical analysis prove the efficiency. To solve the problem that single channel prediction can not efficiently mine the relation hidden in the spectrum data, we use back neural network to do multi-channel combining prediction. We use the spectrum data generated by M/M/N model and the measured data as our simulation data. The simulation result shows that spectrum prediction using multi-channel combine can improve the performance of spectrum prediction.