Spectrum Sensing Based on Graph Weighted Aggregation Operator

Signal processing on graph provides a promising perspective for spectrum sensing. The existing graph-based methods only employ the converted unweighted graph feature but ignore the edge weight and the graph signal information based on the signal-to-graph conversion mechanism. In this letter, we propose a graph-based detector by fully exploiting and merging weighted graph feature and graph signal. In particular, we first propose an improved graph representation framework to convert the power spectrum of the received signal into weighted graph and map that into graph signal. Subsequently, the weighed graph feature of the converted graph is characterized. On this basis, we propose a graph weighted aggregation operator to jointly combine the graph signal and weighted graph feature. Monte Carlo simulation results demonstrate that the proposed method is significantly superior to the existing graph-based methods and energy detection particularly for low signal-to-noise regions.