Study On Cooperative Spectrum Sensing Based On Energy Detection For Cognitive Radio

With the rapid development of wireless communications technology and dramatic in-crease of wireless communications services, licensed spectrum is running out, and unli-censed ISM (Industrial Scientific and Medical) band is also more and more crowded. How-ever, practical measurement shows that:the main reason for current scarcity of spectrum resources is that the low utilization of licensed spectrum under traditional fixed spectrum allocation policy. To improve the utilization of spectrum, ease spectrum scarcity, a variety of dynamic spectrum access technology came into being. Among them, the cognitive ra-dio is considered is perhaps the one with the most potential applications. The key idea of cognitive radio is:when the licensed spectrum band is not being used by primary user, the secondary user could opportunistically access that band without interfering with the primary use; when the primary user re-accesses this band, the secondary user releases the band to avoid conflicting with primary user.Spectrum sensing is a prerequisite challenge for cognitive radio. Recently, collabora-tive spectrum sensing methods have been considered to support reliable spectrum sensing performance, by alleviating the detrimental effects of multipath fading or shadowing, as well as the noise or interference in the wireless environment. In this paper, we focus on the spec-trum sensing for cognitive radio. The main contributions of this dissertation are outlined as follows:1. A simple equal weighted cooperative spectrum sensing algorithm is proposed. And taking into account the effects of multipath fading or shadowing in wireless envi-ronment, the probabilities of detection over Rayleigh fading, Nakagami-m, fading, and Lognormal shadow fading channels are theoretically analyzed, utilizing the characteristics of finite sum of distributed random variables. The simulation re-sults show that the detection performance of proposed method approximate that of N-P optimal rule when SNR is above OdB. Such as, over Nakagami-m channel, the missing probabilities of proposed method and N-P optimal rule are 2.3×10-4 and 2×10-4 respectively as the false-alarm probabilities both are 10-3. However, the computation complexity of the proposed method is much low than that of N-P optimal rule, and the parameters which are utilized are also simpler and less. In practical, the received signal samples need to be quantified. For this situation, the thresholds selection and false alarm probability are analyzed. The result data told that as the quantization bits are as much as 8 bits, the negative effect of quanti-zation is such weak that could be neglected. Such as, the missing probabilities of quantization and non-quantization differ within 2%, as SNR is OdB.2. Comparing with the fixed sample test, sequential probability ratio test often re-quire less samples on average to achieve the same level of error performance. This is important for reducing detection delay causing by interchange local detection results. However, the calculation of probability ratio often is complex for lack of prior knowledge. In this paper, combining the advantages of sequential probability ratio test and the cooperative detection based on hard-decision, a novel sequential differential test (SDT) method is proposed. In this method, we focus on the count difference of local 1 decision and 0 decision, and the final decision is made based on the count difference. The performance of proposed method is analyzed mathe-matical. Numerical results show that SDT method makes a better tradeoff between detection probability and false-alarm probability than OR rule and AND rule. The SDT method can exhibit high detection probability, simultaneously maintain low false-alarm probability. In most SNR conditions, especially as the decision thresh-old p=5, the SDT exhibit higher detection probability and lower false-alarm probability than MAJ rule, i.e., the detection probabilities of SDT and MAJ rule are 50% and 37.5% respectively, and the false-alarm probabilities are 12% and 17% respectively, as SNR is-2dB.3. The spectrum sensing with security of is also considered. In cooperation, some sec-ondary users may become invalid for some reason, and is unable to conduct reliable local spectrum sensing; or, selfish/malicious secondary users will falsify the local sensing reports. Both cases may cause the fusion center to make an incorrect spec-trum sensing decision. Thus, a current weighted sequential probability ratio test (WSPRT) method based on reputation was proposed. However, the local sensing results are binary-decision, that is all local 1 decision (or 0 decision) will be treated indiscriminatingly. For example, a 1 decision made for the test statistics is heavy greater than the threshold and another 1 decision made for slightly greater than the threshold would imply different information. Since the WSPRT method can not distinguish these two types local 1 decision, it means serious loss of local detec-tion information. Therefore, a improved method replacing the binary-decision with multi-bits decision is proposed, to enhanced the ability to resist selfish/malicious user. Moreover, corresponding to multi-bits decision, three types of new reputation evaluating schemes are introduced:neutral, punitive and heavy punitive. We ver-ified that improved method outperforms the WSPRT at the expense of one or two more bits of overhead for each local report. For example, for reputation evaluating scheme of heavy punitive, the missing probabilities of 2-bit and 3-bit soft-decision of proposed enhanced method decrease as much as 20% and 40% respectively.