Optimization Of Spectrum Sensing Techniques In Cognitive Radio

This doctoral dissertation focuses on the noise characteristics and the trade-off optimization between spectrum sensing overhead and the system performances in cognitive radio. Firstly, the noise power uncertainty models are studied for the essence of the SNR walls phenomenon in the energy detection. Secondly, the algorithms to conquer the SNR walls are surveyed comprehensively. Next, the joint optimization of the detection sampling number and the cognitive user number is studied to achieve the trade-off optimization between the overhead and the sensing performance. Lastly, due to the data collisions between the primary and cognitive users, the trade-off joint optimization problem between the cognitive frame parameters and the troughput is reaserched to get the frame optimization entirely. The main research contents and conclusions as follows:1)The source, categories and statistic characteristics of the noise in communication system are studied comprehensively. A log-normal distribution model of the noise power uncertainty is presented. The influences of the noise uncertainty to the sensing algorithm are explored basing on the presented model. It is proved by the confidence interval analysis that the closed interval model of noise power is only one special case of the presented. It will be useful for the further algorithm study of the weak signal detection.2)Due to the fateful impacts of SNR walls to the local sensing algorithms, a comprehensive survey of SNR walls'essence and the algorithm optimizations to solve the problem is presented. The conclusion is that the phenomenon of SNR walls comes from two factors. One is the dependence on the noise power of the decision threshold in the test hypotheses. The other is confusion of the signal and the noise in the test statistic due to the information scarcity of the primary signal. The algorithms to conqur the SNR walls are classified into two sorts. The general optimal sensing algorithm confronting the problem is identified as the one based on the sample covariance matrix eigenvalues.3)The trado-off between cooperative sensing overhead and cognitive network throughput is studied further due to the property of data throughput scaling the spectrum efficiency of the cognitive network. A joint optimization model of the local sampling number and the cooperative user number is presented basing on the soft combination and the throughput maximieren rule. The Armijo inexact line search steepest decent algorithm is modified due to the two lopsided parameters. The global optimal solution of the optimization problem is achieved by the modified algorithm, the optimal combination of the sampling number and user number is achieved. The important parameters in cooperative spectrum sensing are optimized dynamicly. Comparing to the single parameter optimizations, joint optimization reveales the inherent laws more profoundly and solve the problems more thoroughly.4)The data collisions between the primary and cognitive users are modeled basing on the primary business model and the data packet pattern. The validity of the model is proved by Monte Carlo simulation. The effects of the data collisions to the cognitve throughput and the cognitve frame are explored. The trade-off between the sensing cycle and the cognitive network throughput is modeled as an optimization problem and the model validity is proved theoretically. The optimal sensing cycle considering the data collisions is achieved.5)A joint trade-off optimization model of the sensing cycle and the sampling time is presented basing on the cognitive network throughput maximieren rule to solve the data collisions problem thoroughly. Similarly to the method of 3), the global optimal solution of the binary optimization problem is achieved by Armijo inexact line search steepest decent algorithm. The optimal frame parameters configuration is achieved considering the data collisions. Then the problem of data collisions between primary and cognitve users has been solved from the system optimization level to a certain extent.