Resource-efficient fusion with pre-compensated transmissions for cooperative spectrum sensing.
Bottom Line:
The estimates are used at the SUs to pre-compensate for the reporting channel phase rotations and to partially compensate for the channel gains.This partial compensation is the result of signal clipping for peak-to-average power ratio (PAPR) control.We show, analytically and with simulations, that this new scheme can produce large performance improvements, yet reduces the implementation complexity when compared with the original one.
View Article:
PubMed Central - PubMed
Affiliation: National Institute of Telecommunications-Inatel, Av. João de Camargo, 510, 37540-000 Santa Rita do Sapucaí, Brazil. dayan@inatel.br.
ABSTRACT
Recently, a novel fusion scheme for cooperative spectrum sensing was proposed for saving resources in the control channel. Secondary users (SUs) simultaneously report their decisions using binary modulations with the same carrier frequencies. The transmitted symbols add incoherently at the fusion centre (FC), leading to a larger set of symbols in which a subset is associated with the presence of the primary user (PU) signal, and another subset is associated with the absence of such a signal. The decision criterion applied for discriminating these subsets works under the assumption that the channel gains are known at the FC. In this paper, we propose a new simultaneous transmission and decision scheme in which the task of channel estimation is shifted from the FC to the SUs, without the need for feeding-back of the estimates to the FC. The estimates are used at the SUs to pre-compensate for the reporting channel phase rotations and to partially compensate for the channel gains. This partial compensation is the result of signal clipping for peak-to-average power ratio (PAPR) control. We show, analytically and with simulations, that this new scheme can produce large performance improvements, yet reduces the implementation complexity when compared with the original one. No MeSH data available. Related in: MedlinePlus |
Related In:
Results -
Collection
License getmorefigures.php?uid=PMC4482000&req=5
Mentions: In order to allow for an analysis of the accuracy of Equations (21) and (22), Figure 5 shows theoretical and simulated ROC curves of our fusion scheme for M = 3, M = 5 and K = 1 and different clipping thresholds . The received SNR at the SUs was arbitrarily set to ΓSU = −5 dB and the average received SNR per bit at the FC was set to ΓFC = 0 dB. First, we notice that a better accuracy of Equations (21) and (22) is achieved for large values of c, which is an obvious result, since large c means a small probability of clipping and, thus, small probabilities that the noiseless received symbols fall outside their awaited values, i.e., small Pout. However, large values of c lead to transmitted signals with high PAPR, which reduces the average Euclidean distance between the transmitted symbols, eventually reducing the detection performance at the FC. Recall that a high PAPR is the result of the attempt to compensate for low channel gains and that a resulting BPSK signal with high peaks will cause a compression on the original symbol position, so as to maintain the average energy Eb; mathematically, high values of will lead to high values of ξ in Equations (21) and (22). On the other hand, very low clipping thresholds, while good from the perspective of the PAPR, also degrade performance. This is due the fact that a high Pout produces a high influence on the threshold-based decision rule proposed here, because a received symbol may cross the threshold even without noise. Finally, from Figure 5, it is also possible to observe that the procedure described at the end of Subsection 2.2 for determining C is indeed adequate: notice that with c = 2.9 for M = 3 and c = 3.7 for M = 5, i.e., and , respectively, the simulated performance is the best one among those shown, and the gap from the theoretical performance is small. This gap is even smaller in the case of c slightly above the optimum, and the performance is only slightly decreased; however, the PAPR is increased more noticeably, from around 3.2 for c = 2.9, to around 6.8 for c = 5 in the case of M = 3. For M = 5, the PAPR is increased from around 4.5 for c = 3.7, to around nine for c = 6. These values can be easily verified from the curve of C versus PAPR in Figure 2. |
View Article: PubMed Central - PubMed
Affiliation: National Institute of Telecommunications-Inatel, Av. João de Camargo, 510, 37540-000 Santa Rita do Sapucaí, Brazil. dayan@inatel.br.
No MeSH data available.