Graph Learning-Based Cooperative Spectrum Sensing in Cognitive Radio Networks

In cooperative spectrum sensing (CSS), received signal strengths (RSSs) of multiple secondary users (SUs) are combined to improve sensing performance. In existing CSS schemes, RSS levels are often assumed in the same order of magnitude, which, however, is not the case due to, e.g., shadowing in some scenarios, leading to performance degradation. To address this issue, in this letter, we first reveal that the expectation of a RSS matrix from SUs has the property of low rank. Then, using this property, we propose a graph learning method to select SUs of the same order of RSS levels. Specifically, a probability matrix of graph edges is used to represent the correlations of SUs, and a graph learning problem is formulated for learning the probability matrix, which is further solved by using alternating direction method of multipliers (ADMM). Then SUs with high correlations are collected for implementing CSS. Moreover, false-alarm and detection probabilities of the proposed detector are analyzed. Numerical results demonstrate the superiority of the proposed detector compared to state-of-the-art detectors.

Automated summary

In cooperative spectrum sensing (CSS), the received signal strengths (RSSs) of multiple secondary users (SUs) are combined to improve sensing performance. Existing CSS schemes often assume RSS levels to be in the same order of magnitude, leading to performance degradation. To address this issue, this letter proposes a graph learning method that selects SUs with similar order of RSS levels using a probability matrix of graph edges. The proposed detector outperforms state-of-the-art detectors according to numerical results.