Research On Robust Spectrum Sensing For Mutiantenna Receivers

Cognitive radio is one of the potential paradigms for solving the existing spectrumdeficiency. As an important element for cognitive networks, spectrum sensing has re-ceived much attention. The classical non-blind methods rely on the priori knowledge ofthe primary signals, which, however, is usually unavailable in the secondary receiver. Onthe other hand, most of the existing blind approaches are not robust enough in practicalapplications, thereby posing a big challenge to cognitive radio networks. This dissertationaims at enhancing the detection probability and robustness of spectrum sensing methodsin cognitive radio networks and hence is an intensive research topic.This dissertation mainly concentrate on improving the detection models in practicalspectrum sensing based on the communication environments of cognitive radio networks,eventually making them coincidence to practical situations. Based on the so-obtainedmodels, more powerful and robust methods for spectrum sensing are derived. Meanwhile,performance analysis is carried out to theoretically evaluate these methods for practicalimplementations.This dissertation considers some practical spectrum sensing scenarios and, accord-ingly, proposes several algorithms for these cases. For collaborative spectrum sensing,this dissertation first addresses the reason why the background noise is non-uniform inthis scenario and then develops new detection models in the presence of non-uniformnoise. Based on the new model, a new detector is derived in the framework of the GLRT,namely, the Hadamard ratio test. Subsequently, performance analysis for the Hadamardratio test is conducted, ending up with accurate analytic formulae for the false alarm prob-ability and detection probability via the moment matching technique. Numerical resultsare presented to validate our theoretical findings. In particular, the accuracies of the so-obtained distributions are first validated via Monte-Carlo simulation. Then, a comparisonbetween the detection performances of the proposed method and state-of-the-art methodsis performed in terms of receiver operating characteristic. Simulation results indicate thatthe proposed method is superior to the state-of-the-art methods.When there are multi-antenna secondary users in collaborative spectrum sensing,the channels corresponding to antennas of this secondary user are strongly correlated toeach other. To fully exploit this correlated property, a volume-based method is derived by using the hyper-volume of the covariance matrix. The delta-method is then employedto approximately calculate the first two moments of the detector, which are determinedby exploiting the Taylor expansions for the moments of random variable. Numericalresults show that the derived approximations are reasonably accurate under large samplesize. It is also shown that the Hadamard ratio test and volume-based detector have theiradvantages and disadvantages, which depend on the application situations.If the primary signal is non-circular, this important property of the signal should betaken into account in the detection model to further enhance the detection performance.However, although the non-circular signal has been widely used in modern wireless com-munications, its statistical property has not yet been considered in state-of-the-art methodsfor spectrum sensing. In chapter4, a new Hadamard ratio test for non-circular primarysignals is derived in the framework of general likelihood ratio test, which is named asthe NC-HDM algorithm. The NC-HDM approach is able to utilize the property of thenon-circular signals to further enhance the sensing performance. Simulation results showthat the NC-HDM approach outperforms the existing methods when the primary signalshave large non-circularity rates. In practical spectrum sensing, the receivers can switchbetween the NC-HDM approach and other approaches according to the non-circularityrate of the primary signal.