Spectrum Sensing In Cognitive Radio

In today's wireless networks, the effectiveness of the traditional fixed spectrum assignment policy is found being strained by a dramatic increase in the requirement to access to the limited spectrum for mobile services. The inefficiency of the spectrum usage necessitates a new communication paradigm to exploit the existing wireless spectrum in an opportunistic way. The cognitive radio is such the strategy of dynamic spectrum access. It allows the cognitive mobile users to work in the best available channel and extends the bound of spectrum usage efficiently. The cognitive users need to reliably detect signals from licensed users to avoid harmful interference. So the spectrum sensing is a key enabling functionality in cognitive radio networks. In this thesis, we have done the following works:1. The majority of spectrum sensing algorithms assumed that all collaborative cognitive radio users have the same threshold without considering the influence brought by different environment. Under analysis of the main elements affecting the performance of spectrum sensing,we propose a spectrum sensing method based on energy detection. The cognitive user with low signal-to-noise rate searches a collaborative user constrained by some conditions. With a constant joint probability of false alarm, cognitive radio users adjust their thresholds according to SNRs to increase the high-SNR user's contributions to the performance of collaborative spectrum sensing and constrain the negative effects caused by low SNR user. The judgment results can be transmitted by using one bit which can decrease the communication costs.2. An improved collaborative spectrum sensing algorithm was proposed based on likelihood ratio. We enhance the performance of spectrum sensing from two points, local sensing and global decision. A linear framework is adapted to arrange different contributions of samples to improve the performance of local spectrum sensing. According to central limit theorem, an approximately likelihood rate test, in the case of a large number of cooperating cognitive users, is developed to improve the performance of global decision.3. In order to achieve a good tradeoff between sensing performance and complexity, we proposed a Multi-threshold scheme with optimized weights. Different weights are arranged to corresponding different intervals to indicate their reliability. We also gave the method to compute the optimal weights theoretically. The optimal user number for each interval was estimated based on Minimum Error Probability, and then the optimal weights were computed.