LLR based spatiotemporal cooperative spectrum sensing schemes for cognitive radios

Most existing distributed soft combining spectrum sensing usually assume that all sensing users either have the same signal-to-noise ratio (SNR) values of primary user signal, or are independent of each other. But these assumptions are not always applicable. In this paper, a distributed soft combing spectrum sensing scenario with correlated cooperation users is considered. The secondary users may have unequal SNR values. In order to alleviate the well-known challenge of large amount of data reporting, two simple but effective local test statistics derivation schemes by averaging power samples or logarithmic likelihood ratios (LLRs) are proposed. Instead of sending all raw data, cooperation users just send their finally averaged LLR values to the user in need. In this way, the decision user can avoid estimating the covariance matrix as well as the eigenvalue decomposition, and omit the reliance of prior joint distribution of received signals, which is hard to get in this scenario. Furthermore, by exploiting spatial sensing results, only part of users selected according to three kinds of user selection rules rather than all need to send their LLRs. The decision user adopts equal gain combining (EGC) or LLR-weighted fusion to make the final decision. Simulations are carried out in three different correlation scenarios to show the detection performance of proposed cooperative spectrum sensing schemes.