On Spectrum Sensing In Cognitive Radio

With the rapid development of technologies and applications in wireless com-munications, wireless service is focusing on those applications which require morebandwidth and higher rates, such as multimedia communication. But the lack of ra-dio spectrum is becoming the obstacle of these applications. In fact, the spectrumshortage is mainly due to the old spectrum allocation policy but not the lack of spec-trum itself. Under the current spectrum management framework with fixed allocation,most spectrums are found to be underutilized. Cognitive radio can dynamically reusethose idle spectrum resources in temporal and spatial domain, and provide frequencybands for new wireless applications. Thus, it is considered to be one of the promisingtechnologies to achieve dynamic spectrum access and solve the spectrum shortageproblem in the future. Cognitive radio offers a variety of research fields, including spec-trum sensing, sharing and management. This dissertation deals with the problem ofspectrum sensing in cognitive radio. The main contributions are as follows.1. The problem of primary signal detection under noise uncertainty is considered.A detection method with noise calibration is proposed, which can reduce theeffect of noise uncertainty on the detection robustness by estimating the noisepower in the guard band before detection. The selection of noise calibrationduration under the restriction of robustness is deduced. Moreover, the determi-nation of the parameters to maximize the detection performance in quiet periodis given. With known noise uncertainty, a cooperative sensing method with dual-threshold is proposed. The thresholds are determined by the noise uncertaintyat each secondary user. Simulation indicates that the proposed method outper-forms the traditional methods in detection performance.2. The use of multiple antennas in wireless communication can provide spatialdiversity and improve the reliability of transmission. In this paper, based onthe fading among antennas, a spatiotemporal sensing method over Rayleighfading is proposed. By performing GLRT(Generalized Likelihood Ratio Test)based on the fading among antennas and the variance in primary signal samplesseparately, the primary signal can be detected without the knowledge of thenoise power. So this method is robust to the noise uncertainty.3. During the spectrum sensing, quiet period which ceases the on-going trans- mission is used to avoid the interference of self-signal with the detection. Theinterruption of transmission in the quiet periods brings additional overheads andaffects the transmission performance. With the purpose of reducing these over-heads, the primary detection method without quiet period is investigated. Adetection method with linear self-signal elimination based on the characteristicsof fading channel is proposed. Focusing on the OFDM(Orthogonal FrequencyDivision Multiplexing) system, a low complexity detection method with comple-mentary symbol couples is proposed. It requires no quiet period but have a sim-ilar performance with the traditional energy detection with quiet period. Finally,with the given self-signal elimination model in OFDM, the performance of someexisting methods based on self-signal elimination are analyzed and evaluated.4. Influenced by some factors, such as the fading and noise uncertainty, detec-tion with a single detector will not give reliable results. It is an effective mannerto improve the detection performance by cooperative sensing among multipledetectors. However, the messages exchange in the collaboration brings lotsof overheads. In this paper, a cooperative sensing scheme is proposed basedon coherent MAC(Multiple Access Control) and analog forwarding. The over-heads are reduced from O(n) to O(1). With the analysis of the performanceoptimization, a low complexity weighted coefficient allocation method under theconstraints of interference to primary users is proposed. Compared with thedecision fusion with CDMA(Code-Division Multiple Access), an orthogonal MACwhich has similar overheads, our method can greatly improve the detection per-formance.