Abstract :

Cognitive Radio is a promising technology to resolve spectrum scarcity issue by exploiting RF spectrum in opportunistic fashion. Spectrum Sensing is a key enabling step towards successful implementation of this emerging technology. Sensing refers to the detection of unused spectrum spaces also known as "white spaces". This dissertation proposes, investigates and analyses several algorithms for spectrum sensing cognitive radio applications. Receiver Operating Characteristic (ROC) is compared among different proposed algorithms. It is one of the most important measures to classify detectors. In the first part of the dissertation, ROC is analysed for multiple-antenna assisted spectrum sensing radios under shadowing and is derived using linear test statistic. Furthermore, a highly useful cluster-driven architecture for spectrum sensing is also proposed and analysed that improves detection probability by exploiting cooperation among cognitive radios using hard decision combining strategy. Hard decision combination strategy computes the detection probability by combining one bit decisions among various cooperative cognitive radios. Detection probability is achieved 80% at PFA rate of 10% for a single user, whereas using hard decision combing approach the same detection probability is achieved at 1% P. For Binary Symmetric Channel with 10 error probability, P results 32% (at P,, IO) for a single user whereas 65% for a five user case. In the second part of the dissertation, a novel channel model ie. double exponential correlation is incorporated for spectrum sensing algorithms under suburban environments. Asymptotic probability of detection is derived, analysed and compared with classical exponential correlation model (also known as Gudmundson's model). Using proposed model missed detection probability reaches Zero for less than ten sensors whereas Gudmundson's model results a constant 0. 7 missed detection probability even when the sensing users are increased to hundred. Thus, results verify that the proposed model performs significantly better than the classical one. In the third part of the dissertation, a cooperative sensing strategy is proposed for mobility-driven cognitive radio. It is also verified through simulation results that the proposed decision-fusion based architecture performs significantly better than the independent sensing radios. Using collaboration under urban environments, missed detection results in 30% in comparison to 62% for a single user under false alarm probability of 10%.