| dc.description.abstract | In the past few years, there have been remarkable developments in wireless communication technology, leading to a rapid growth in wireless applications. However, this dramatic increase in wireless applications is severely limited by bandwidth scarcity; a fundamental resource for communications. Traditionally, fixed spectrum assignments are used in which frequency bands are statically assigned to licensed users. The static spectrum allocation fails to provide vacant spectrum bands to new coming users and services. Hence, a new communications and networking paradigm based on dynamic spectrum allocation has emerged, namely cognitive radio system. In cognitive radio networks, spectral utilization is improved by allowing unauthorized (secondary) users to regularly sense the radio spectrum and opportunistically use frequency bands not utilized by licensed (primary) users. Primary users have higher priority than secondary users; therefore, secondary users need to utilize idle spectrum holes without causing harmful interference to primary users. In order to achieve minimum level of interference to primary users, efficient spectrum sensing techniques need to be implemented. Spectrum sensing is one of the main challenges in opportunistic spectrum usage, since it is responsible for providing efficient and fair spectrum access and scheduling among licensed and unlicensed users. Cooperation between cognitive radio users has been proposed in the literature to overcome spectrum sensing challenges by providing spatial diversity. Both centralized and decentralized cooperative spectrum sensing systems have been implemented to enhance detection capability of cognitive radio networks. In this work, we formulate the cooperative spectrum sensing process as a pattern recognition problem, where a centralized node classifies the target spectrum into two classes: busy (presence of signal) or vacant (absence of signal). The classifier is designed to identify white spaces in the spectrum while minimizing interference with licensed primary users and maximizing spectral utilization in environments exhibiting shadowing and fading effects. Polynomial classifiers were proposed in this work as classifier models, in which first and second order expansions are investigated. Feature extraction stage consists of two spectrum sensing techniques: parametric and non-parametric. In nonparametric detection algorithms, such as energy detection and autocorrelation, the cognitive network does not have a priori knowledge on the primary users' signals. On the other hand, in parametric detection, cyclic features characterizing primary signals and prior knowledge of synchronizing preamble patterns are utilized. The parametric detection schemes include coherent detection and cyclostationary feature detection. Extensive simulations were performed to design, model, and evaluate the cooperative classifier system, when both parametric and nonparametric features are used. In nonparametric spectrum sensing, simulations demonstrate the superior performance of autocorrelation detection scheme over energy detection. Moreover, simulations of parametric spectrum sensing indicate that cyclostationary feature detection outperforms coherent detection. Finally, it was shown that parametric sensing schemes yields a superior performance over nonparametric sensing schemes when implemented under same conditions. | en_US |
