The Research Of Performance Improvement Methods For Cooperative Spectrum Sensing In Cognitive Radio

As one of the most promising dynamic spectrum access technologies, cognitive radio (CR) technology can solve the conflict between the shortages of spectrum resource and underused licensed spectrum bands. The core idea of CR technology is sharing licensed spectrum without interfering with primary users. Thus one of the most important functions of CR is to detect idle spectrum bands in its radio environment, i.e., spectrum sensing technology. The essence of spectrum sensing is differentiating the signal of primary users from noise. In general, there is no prior knowledge of the signal and locations of primary users. On the other hand, primary users will not share their licensed spectrum with cognitive users with inferior sensing performance. For above factors, conventional signal detection methods can not be applied in CR systems directly.With the cooperation of a few cognitive nodes at different locations, cooperative spectrum sensing can achieve spatial diversity of signal sources, improve the sensing performance significantly and become one of the most promising spectrum sensing methods in CR. In the dissertation, we mainly study cooperative spectrum sensing technology in CR systems. To improve the sensing performance, some existing important questions and challenges of cooperative spectrum sensing in signal processing, transmission, and data fusion are studied. The major contributions of the dissertation include:(1) Based upon the analysis of the relationship between the energy detection algorithm and signal power spectral density (PSD) function, we develop a correlation-based detection method. According to the relationship between the signal correlation and the PSD function, the central spectrum component of PSD function is adopted as the observation statistics to improve the sensing performance. To compromise the deficiency of the proposed method for uncorrelated signals, an adaptive spectrum sensing model is proposed. In the model, the correlation of received signal samples is estimated and appropriate detection scheme is selected adaptively with trivial increase of computational complexity.(2) Considering the difference of cognitive nodes in sensing performance, a cooperative nodes selection scheme is proposed based on the individual characteristic on the analysis of the relationship of sensing performance between cooperation entirety and the single node. The optimal sensing performance problem for cooperative spectrum sensing under limited time is formulated as a binary nonlinear integer programming (BNIP) problem. The binary particle swarm optimization (BPSO) algorithm is adopted to obtain suboptimal solutions of cooperative nodes. For different system demand, two different optimal objective functions are considered:optimal sensing performance (minimizing the average Bayesian detection risk) and maximizing the throughput of CR system. Theoretical analysis and computer simulations verify that the proposed schemes can achieve less average Bayesian detection risk or higher throughput than the case that all neighboring cognitive nodes participate in cooperative sensing without discrimination under different scenarios.(3) In heterogeneous network, the mobility of cognitive nodes will lead to fluctuation of the received signals power and noise power. Traditional static local sensing scheme with fixed threshold can not keep working at the optimal sensing state at any time. To solve such a problem, an adaptive threshold scheme for cooperative spectrum sensing is proposed. Employing the Steepest Descent Algorithm (SDA), all cooperative nodes adjust their thresholds according to their history sensing performance and the optimal data fusion rule is adopted in the control center to decrease the average Bayesian risk. No prior information of primary signals, channel fading and noise power is needed and the optimal sensing performance is achieved with the proposed scheme.(4) Comparing with non-cooperative spectrum sensing methods, cooperative spectrum sensing can improve the sensing performance, but introduce additional power consumption and reporting delay. Based on the analysis of power consumption model of cooperative spectrum sensing, we propose an energy efficient cooperative spectrum sensing schemes. First, all cooperative nodes are separated into several clusters according to their locations. The energy consumption of the nodes far away from the control center for reporting local decisions is decreased greatly because their communication distance is decreased. Then the binary sequences of local decisions are coded by run-length-encoding (RLE) technique before reporting. Thus some reporting energy consumption is saved through reducing the number of reporting decisions.