The Research Of Spectrum Sensing Algorithms With Low Complexity Based On Mathematical Statistics

Cognitive radio (CR) is considered a kind of wireless communication concept which can be self-cognition and self-learning. It can be used to detect the spectrum resources of the surrounding environment in real time through a series of methods, find "spectrum holes" and improve spectrum utilization. Spectrum sensing (SS) technology is one of the key technologies of cognitive radio. It is the premise and foundation of the key technology of cognitive radio, such as dynamic spectrum access and spectrum coexistence. SS is used to monitor the spectrum state. When primary user (PU) not use the interesting radio spectrum, the sensing user(SU) can access the spectrum band without interfering with the PU communication. When PU uses the spectrum, sensing user (SU) should withdraw from the spectrum instantly to minimize the interference. Thus, SS is the key technology to the development of CR applications. Finally, the work of this thesis are summarized, and the future research contents are discussed.Firstly, in this thesis the research background and present research states of the subject are briefly presented, then the detection model of SS algorithm are introduced, the current spectrum detection algorithms are briefly presented in classification? Three kinds of spectrum sensing algorithms based on goodness of fit (GOF) are introduced, for example AD algorithm, blind spectrum detection algorithm based on T distribution and blind spectrum sensing based characteristic function and Anderson-Darling test(CAD). Meanwhile, MATLAB simulation experiment is given.The last two kinds of improved AD algorithms do not require any prior knowledge. In slow fading Rayleigh channel, the two improved AD algorithm and AD algorithm which the variance of noise was known have similar detection performance, especially under the condition of small sample. At the same time, the detection performance of the three algorithms based on the goodness of fit algorithm is significantly higher than that of the ED algorithm.Secondly, aim at the shortcoming that the current GOF algorithm has high time complexity, two new low time complexity spectrum sensing algorithm was proposed by using the method of mathematical statistics from the perspective of classical statistics and Bayesian statistics. For example spectrum sensing based on the generalized likelihood ratio and blind spectrum sensing based on Bayesian Inference. The detection performance of spectrum sensing based on the generalized likelihood ratio is slightly higher than that of the AD algorithm's detection performance and has low complexity. The detection performance of blind spectrum sensing based on Bayesian Inference is similar to that of the CAD algorithm's detection performance. However, its time complexity is significantly lower than CAD algorithm and can overcome the influence of noise uncertainty.Finally, based on the idea of modularization, a static signal spectrum sensing simulator is designed by using the MATLAB GUI tools. This simulator has a good expansibility and can achieve the simulation of spectrum sensing algorithm quickly. Meanwhile, the research work of this thesis is summarized, and the future research direction is prospected. SS technology as a key and basic technology for CR has more application prospect and research value.