Noncooperative Spectrum Sensing with Historical Sensing Data Mining in Cognitive Radio

In cognitive radio (CR), spectrum sensing is a critical task to effectively detect the spectrum holes. Through dynamically accessing such idle spectrum, CR can improve spectrum utilization and provide broad wireless band for wireless services. There are two types of spectrum sensing algorithms for CR, which are cooperative and noncooperative spectrum sensing. Noncooperative spectrum sensing is easier to implement, has less power costs, and is more robust to node failure, compared with cooperative spectrum sensing. In this paper, a novel noncooperative spectrum sensing algorithm has been proposed to exploit historical sensing data and boost the spectrum sensing performance. The hierarchical Dirichlet process is used to cluster the historical sensing data into groups, where each group shares a consensus spectrum state. Through exploiting the statistical correlations among historical sensing data, the uncertainty in spectrum sensing can be significantly minimized. The simulation results show that the proposed algorithm is better than the other three typical algorithms and has more than 10% detection performance improvement at false alarm probability 0.2 when received signal-to-noise ratio (SNR) is 5 dB.