The Study Of The Cooperative Spectrum Sensingalgorithm Based On Data Fusion

Recently, intense competition for spectrum usage has arisen, and Cognitive Radio (CR) has become a hot spot. A CR network composed of CR users can coexist with primary users under the dynamic spectrum access paradigm. A Secondary User (SU) should be able to scan through spectrum bands and find vacant bands to operate in. To avoid interfering with Primary User (PU), a SU needs to carry out accurate spectrum sensing, which becomes a key technology. Cooperative spectrum sensing enables a Cognitive Radio (CR) networks to detect primary users much more reliably comparing with the uncooperative spectrum sensing. And, the data fusion technique is a key component of cooperative spectrum sensing.Firstly, the thesis researches the cooperative spectrum sensing methods based on energy detection, which include hard combination detection and soft combination detection. Hard combination detection contains"OR"rule,"AND"rule and vote fusion rule. Soft combination detection contains the largest posterior probability rule, Bayesian rule, N-P rule, EGC rule, MRC rule and zoning rule. Through the MATLAB simulation, if the probability of false alarm in"OR"rule and"AND"rule is constant, there is a optimal number of SU which can make the probability of detection biggest. Vote fusion rule's optimal threshold should be around half cognitive radio numbers. And, The performance of maximal ratio combination is better than equal gain combination. The largest posterior probability rule can approach ascendant sensing performances than"OR"rule and"AND"rule.Then, two zoning rules are researched. One is combined EGC and MRC rule into censoring scheme of zoning rule, the other is decision fusion based on zoning rule. In decision fusion rule, the SU will detect again, when its energy falls into the second region. For decision rule, its performance of spectrum sensing is investigated for both perfect and imperfect reporting channels, and the close formulations of the detection probability are presented. And we mainly investigate two imperfect reporting channels, that is Rayleigh imperfect channel and Nakagami imperfect channel. The MATLAB simulation results show that the probability of error is biggest and the probability of detection is worst by Rayleigh channel. With the parameter g increases, the signal fading leveling off caused by Nakagami channel. At last, the thesis mainly researches the largest posterior probability rule with quantization.The censoring schemes combined with the locally optimal quantization and the uniform quantization for cooperative spectrum sensing are proposed to increase sensing bits to the fusion center, and get the better spectrum sensing accurate. The thesis investigates the performance of the largest posterior probability rule with single threshold and double threshold for both perfect and imperfect reporting channels. CR dose not send the result into fusion center, when its energy falls into middle region, by the largest posterior probability rule with double threshold rule. Taking two regions quantization and three regions quantization for an example, we propose a new rule which combines hard combination and the largest posterior probability rule, which not only improves detection performance but also reduce the complexity of the system. In this rule, the fusion center combines the hard decision and the result of the largest posterior probability rule. Simulation results show that the performance of the locally optimal quantization is better than the uniform quantization in the CR system. And the spectrum sensing performance of the new rule is good when SNR is very smaller.