Cyclostationary Feature Based Quickest Spectrum Sensing in Cognitive Radio Systems

Cyclostationary feature is incorporated into the framework of quickest spectrum sensing, which is based on the technique of quickest change detection, in cognitive radio systems. Different from detecting the change in a sequence of independent random variables, e.g. energy detection, the property change of random process with correlated observations should be detected in the context of cyclostationary feature based spectrum sensing. Motivated by traditional cumulative sum (CUSUM) test in quickest change detection, a multi-thread competition algorithm for detecting the change in cyclostationary feature is proposed to tackle the difficulty of correlated observations. A thread truncation approach is proposed to reduce the required amount of memory and computation. Numerical results show that the cyclostationary feature based quickest spectrum sensing significantly outperforms the traditional block detection based spectrum sensing scheme.