Research On Wide-band Spectrum Sensing Based On Compressive Sensing

Developing based on software-defined radio, cognitive radio is an intelligent radio technology which is able to discover and utilize idle authorized spectrum which is allocated to primary user but not being used temporarily, via sensing external wireless spectrum environment persistently. Cognitive radio can efficiently solve the growing problem of spectrum resource scarcity caused by the rapid development of wireless communication technology and fast growth of wireless users. As the core technology of constructing and implementing cognitive radio, spectrum sensing enables cognitive radio system to discover and utilize spectrum resource, while protecting authorized primary user from harmful interference. As the rapid increase of wireless services and growing need of wireless spectrum resource, in order to improve the utilization of spectrum resource, wide-band spectrum sensing has become an important research direction in the field of cognitive radio.In wide-band cognitive radio systems, since the utilization of wide-band spectrum is very low, and primary users only occupy a small amount of spectrum, the signal in frequency domain has sparsity. Compressive spectrum sensing utilizes such sparsity, projects high-dimensional signals on low-dimensional measurements, and reconstructs original signals using optimization algorithms. In this way, wide-band spectrum can be scanned directly at sub-Nyquist rates, and efficient wide-band spectrum sensing can be realized. However, wide-band compressive spectrum sensing encounters many challenges: First, hidden terminal problem is brought by wireless fading environment, and harms the detection performance of wide-band spectrum sensing; Second, because of the lack of prior knowledge on sparsity order, cognitive users will have difficulties in deciding accurate sampling rates while sensing the wide-band spectrum, which will lead to sampling wastage or poor sensing performance; In addition, in the cognitive radio system there may exist some malicious cognitive users, who will attack the system by sending false spectrum sensing data during the data fusion process, which will impair the performance of wide-band spectrum sensing seriously.In order to solve above problems, this dissertation research on how to improve the performance of wide-band spectrum sensing using compressive sensing theory and collaborative spectrum sensing technology, in both centralized and distributed cognitive radio networks. Based on sequential compressive sensing, a wide-band spectrum sensing algorithm which can adaptively determine the optimal number of random measurements is proposed to solve the problem of sensing without the prior knowledge of the sparsity order. Additionally, a reliable wide-band compressive spectrum sensing is proposed to defend malicious users. The main work and contributions of this dissertation are as follows:Firstly, this dissertation introduces traditional narrow-band spectrum sensing techniques, and analyzes the challenges faced when sensing wide-band spectrum. Compressive sensing theory is introduced, and the technical characteristics of wide-band spectrum sensing are analyzed. Then collaborative spectrum sensing technology, which can solve the hidden terminal problem, is introduced. This dissertation has proposed a detailed analysis on the structures of multi-user centralized and distributed wide-band cognitive radio networks, and provided specifications to evaluate the performance of wide-band compressive spectrum sensing.Secondly, in order to solve the hidden terminal problem and reduce the burden of high sampling rates when sensing wide-band spectrum, this dissertation has proposed both centralized and distributed wide-band compressive spectrum sensing algorithms respectively for centralized and distributed cognitive networks. Cognitive users sample wide-band signals at sub-Nyquist rates, and gain spatial diversity gain via collaboration to relieve the negative influence caused by fading environment. In addition, original signals are reconstructed utilizing joint sparse property via iterative support detection. Simulations show that the proposed algorithms can reduce the sampling burden, and have better wide-band detection performances than conventional compressive spectrum sensing methods.Thirdly, in the process of wide-band compressive spectrum sensing, the lack of prior knowledge of wide-band spectrum sparsity order will cause the system to employ overmuch or inadequate random measurements, which leads to sampling wastage or poor sensing performance. To solve this problem, this dissertation has proposed a novel wide-band compressive spectrum sensing algorithm based on sequential compressive sensing and sparsity adaptive matching pursuit. By obtaining sequential random measurements and reconstruction error, minimal number of random measurements can be determined. Simulations show that the proposed algorithm can utilize system resource efficiently, and achieve the desired spectrum sensing performance while avoiding the sampling wastage.Last but not the least, due to the openness and configurability, cognitive radio system may encounter spectrum sensing data falsification attacks by malicious cognitive users. To solve this problem, common forms of attack are analyzed. A reputation-based algorithm is proposed for centralized cognitive radio networks, where fusion center uses the reputation of cognitive users to determine their weights in data fusion, and eliminates the negative influence caused by malicious users. As for distributed cognitive radio networks, a consensus-based wide-band compressive spectrum sensing algorithm using dynamic threshold to adjust fusion weights is proposed. Simulations show that the proposed algorithms can both effectively reduce sampling costs, successfully combat spectrum sensing data falsify attacks, and achieve accurate and reliable wide-band spectrum sensing in respectively centralized and distributed cognitive networks.