On The Research Of Cooperative Spectrum Sensing For Cognitive Radio

Spectrum Sensing is the key technology for Cognitive Radio (CR) because it is crucial for the success of cognitive radio and it will guarantee service improvement to unlicensed users and avoidance of interference to licensed users. This research is supported by projects like National Science Funds, Beijing Municipal Natural Science Fund etc, and attempts to make a contribution to the theory and application of spectrum sensing for the system cognitive radio network.This dissertation mainly concentrate on the increase of reliability and efficience of cooperative spectrum sensing by multiple cognitive radio users, and studies the fading, shadowing effects and the hidden terminal problem on the performance of spectrum sensing. The main contributions and creative algorithms which are developed in this dissertation are listed in the following:We make an exploration into the spectrum sensing from the view of distributed detection and attempt to increase the detection probability and to lower false alarm probability for the whole cognitive radio network and examine the detection performance of spectrum sensing under different network configuration. When the spectrum sensing nodes are similar with each other i.e. their observations are from same probability distribution, the optimum fusion network is in parallel configuration. However the detection performance of parallel configuration decreases greatly when the detection nodes are dissimilar due to fading and shadowing of radio channels especially when most of them are at low SNR states its performance is inferior to that of high SNR node. The hierarchy fusion scheme we proposed in this paper incorporates both parallel and serial configuration and classifies detection nodes into different groups by their SNR levels. The receiver operation characteristic (ROC) of hierarchy fusion rule shows that the detection efficiency and reliability can be increased no matter the detection nodes are similar or dissimilar.We also analyze fusion rules and deduce their properties and conditions because at present only "OR" rule and "AND" rule are considered and the influence of the node number, the false alarm and detection probabilities are disregarded for most research work. Actually these properties and conditions can be used to improve the detection performance as a low complexity method and the uncertainty of spectrum sensing due to the lack of aprior knowledge of licensed signals can be lessened when proper fusion rule is chosen.When communication is under bandwidth constraint, spectrum sensing sensor nodes normally only send one bit report to fusion center and censoring detection method is employed by sensor nodes. However only conventional fusion rules are considered and the "No decision" information by censoring sensor nodes is disregarded by recent research work. The censoring fusion scheme we proposed incorporates "No decision" message for decision fusion can significantly outdo conventional fusion rules on the improvement of the detection performance of spectrum sensing without traffic increase. By analytical results and numerical computation, we demonstrate that the censoring fusion scheme can make great improvement on the detection performance of spectrum sensing even thoug the node number is small.At last part we attempt to apply decision fusion rules into the security issue of spectrum sensing and propose a decision fusion scheme that combines with the outlier detection method. This scheme can improve the robustness of spectrum sensing when some sensing nodes are malfunctioned or malicious and make a new attempt to CR security issue from distributive detection aspect.